Since 494fb56 and 4c8e4a0, fetchStars wrote stars.bin, stars-meta.bin and stars-index.json, and
fetchExoplanets wrote them again with the 3 277 hosts it adds from the archive and the 574 Gaia
designations it renames, after the live Horizons and archive fetches in between. A run stopped
between the two writes, or fetchStars.ts run on its own as the README documented, left 452 294
stars on disk with no archive star and 574 hosts back to their designations, and the existing
exoplanets.json pointing 4 237 planets at stars that were not there. It happened on this branch:
the dev server served those files.
fetchStars now returns the stars and writes nothing; fetchExoplanets, which build.ts runs after
it, is the only writer. fetchStars.ts run on its own runs fetchExoplanets, which fetches the
stars itself. Checked from cache: `tsx tools/etl/fetchStars.ts` exits 0 and writes 455 571 stars,
the published files byte for byte; build.ts with fetchSolarSystem made to throw fails with
"simulated Horizons outage" and leaves every data file unchanged, where before it left the
452 294-star catalogue. The README's table now names fetchExoplanets.ts as what writes them.
The ETL has no unit harness; these two runs are the check.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
4c8e4a0 added 3 277 hosts from the archive, 2 883 of them past 250 pc and 916 past a kiloparsec,
most in the Kepler field. Every host got a 12 px ring whatever its distance and a place in the
draw budget's first tier after the pinned stars, rules set when 1 of 1 379 hosts lay past 250 pc.
At boot the opening view held 3 332 rings against 1 290 before, 1 687 of them in its lower right
(x > 1000, y > 550 at 1600x1000) where there were 116, over the 250-350 pc grid labels; hosts 1-3
kpc away read as neighbours in a view whose range is 307 pc, and 629 stars past a kiloparsec were
drawn where there had been 3.
Rings and the host tier now take the hosts within SURVEY_EDGE_PC only. The far hosts still compete
for the budget by brightness, and search still enters them. Measured in the running app at
1600x1000: 1 853 rings in all, 1 698 in the opening view, 199 in its lower right; 31 stars past a
kiloparsec drawn; 1 767 hosts drawn. This reverses "a ring on every host" for the far ones: that
was the user's rule of 2026-09-17, when the only far host was one star, so it is flagged for them.
The scene test's fixture gains the lens's planet, 6.7 kpc out. The new test checks the ring count
and the host tier; the opening-view test now expects the lens out of that tier. Control: dropping
the distance condition fails the new test and that one (2 of 824).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
d1aa22e framed a giant so its disc took half the view's tighter half-extent, inside the ring its
neighbours are named on at 0.74 of it. It measured that at 1600x1000 only. The names hang a fixed
distance in from the ring, 74-78 px, so on a 390x844 phone the 0.24 of the half-side between disc
and ring is 47 px and the names land on the disc: Betelgeuse was drawn 98 px wide with HD 39374's
name 70 px from its centre, Antares with HD 148199's at 68 px.
The framing now takes the canvas's shorter side and keeps the disc inside the ring less 90 px,
at most half the half-extent as before, at least a tenth. Measured in the running app at 390x844
on arrival: Betelgeuse, Antares and Rigel drawn 54 px wide, the nearest names at 70, 68 and 127
px. At 1600x1000 nothing changes: Betelgeuse 250 px, its nearest name at 296. The ring's fraction
moves to system-framing.ts, beside the rule that depends on it, and the scene reads it there.
The review's other half, that the Readout sheet covers 44 % of the disc once opened on a phone, is
not changed: the dock opens with no panel below 640 px and folds it on any tap on the scene, so
the star arrives with nothing over it.
Control: giving the names no reach fails "keeps a giant's disc clear of its neighbours' names on
a phone" alone. The scene's passing of the canvas size has no unit test (the test canvas has no
size); the app measurement above covers it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
23547de gave archive hosts measured in no colour a B-V read off the dwarf sequence at their st_teff,
and the card's Colour row says so, but the subtitle then estimated a type from that colour and
said "from colour". 30 of the 54 flagged stars read that way: PSR J1719-1438, a millisecond pulsar
whose only input is st_teff 4 500 K, read "Spectral type ~K5, from colour" beside "B-V 1.13, from
its temperature"; DP Leo ~B7, ZTF J1828+2308 ~B5, ZTF J1230-2655 ~A0.
The subtitle now ends ", from its temperature" for those 30. In the running app the pulsar's card
reads "Spectral type ~K5, from its temperature" above "B-V 1.13, from its temperature". The type
itself is still the dwarf the temperature matches, which a pulsar is not; this only stops it
naming a measurement that does not exist.
Control: ignoring the flag fails "says an estimate came from the temperature where the colour
was read off it" alone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Once the archive's 3 277 hosts became stars and 574 Gaia hosts took their names, "K2-18" listed
the star K2-18 and then K2-180 to K2-186, and no planet: "k2-18 b" and "k2-180" were both prefix
matches, and on a tie the ranking put every star before every exoplanet. Replaying each host's
name through the ranking over the published catalogue, 325 hosts had some of their own planets
pushed out of the eight rows shown (655 planet rows), against 75 before the archive stars.
A prefix match that ends where a word does now scores 3.5, between an exact match and a prefix
running on into the same word. The same replay gives 50 hosts and 62 rows. In the running app
"K2-18" lists K2-18, K2-18 b, K2-18 c, then K2-180 onward; "Kepler-186" lists the star and
Kepler-186 b to f before Kepler-1860; Kepler-22 b and WASP-12 b are back. "Proxima", "Io" and
"TRAPPIST-1" list as before.
Control: scoring the word-ending prefix level with any prefix fails "lists a host's own planets
ahead of the systems whose names run on from its own" alone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2242fe0 gave every star the Sun's photograph in grey, tinted at its temperature and limb-darkened
as 1 - 0.6(1 - mu), and nothing tested any of the three: with the coefficient at 0, the tint
dropped from the material, or the photograph replaced by the tint alone, all 817 tests passed. The
one colour test read scene.starTint, a uniform the scene sets whether or not the material uses it.
The new test enters Proxima and walks the disc material's colour node graph for the tint uniform,
the texture loaded from SUN_TEXTURE_PATH and the coefficient 0.6. Controls, each failing it alone
(1 of 820): the coefficient at 0; at -0.6, a limb brighter than the centre; the tint dropped; the
photograph replaced by the tint. A mutant that keeps the texture node in the graph but multiplies
it by nothing is not caught; the test checks the parts, not the arithmetic between them.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The star field reads every star's temperature at boot, and each read asked giantSurface, which
splits the type and runs two regular expressions whether or not the star is a giant: 455 571
times for 2 888 distinct type strings. It depends on the type alone, so it is now kept per type.
The tint pass over the published catalogue gives the same colours (checksum 5 830 745.929 before
and after) and in Node takes 167-237 ms against 241-305. In the running app, the star field
rebuilt in the page three times after each of four cold boots: median 65 ms before (60-79), 51
and 53 ms after in two runs (46-67). The longest task after the data lands did not move beyond
the noise (median 851 ms before, 859 and 827 after): most of the boot growth since c38a42c is the
larger catalogue and the per-star colour lookup, which the tint needs. Memoising colorIndexToRgb
on its three inputs, the fix the review also named, would not help: 337 998 of the 455 571
triples are distinct, and reviewers measured it 260-370 ms slower.
No behaviour changes, so no new test; the suite (819) passes as before.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
95ffb00 said the field took "the same function and the same temperature the disc is drawn with".
It did for stars without planets, but a host's disc is drawn at the archive's st_teff
(starSurfaceOf), and the field read the host's temperature off its colour. Of the 4 485 hosts with
a temperature on the published catalogue, 1 755 were more than 0.1 in RGB from their own disc:
Kepler-186 at 3 096 K in the field and 3 788 K on its disc, HD 97048 at 6 825 and 10 000.
The scene now hands StarFieldRenderer each host's first published temperature among its planets,
the rule starSurfaceOf uses (publishedTemperaturesK, beside it), and the field tints that star at
it. Over the catalogue the hosts more than 0.1 off their disc go from 1 755 to 0. Measured in the
running app at 1600x1000, field tint against the disc's starTint after entering each: Kepler-186
(1, 0.620, 0.326) against (1, 0.619, 0.326), where the field was (1, 0.498, 0.174); Teegarden's
Star 0.001 apart; Proxima, HD 97048 and Sirius 0.
The field's test compared colorIndexToRgb with the same formula. The new scene test enters Proxima,
whose B-V 1.8 reads 3 070 K and whose disc is drawn at the archive's 2 900, and compares its field
colour with the disc's. Controls, each failing that test alone: the field ignoring the map it is
given, and the scene passing an empty one.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
giantSurface carried the 64-71.5 fit of van Belle et al. (2021, ApJ 922, 163, table 8) on to
index 73.9, where the table's fourth piece, 72-74, is flat at 3 134 K: M6 III (index 72) came out
3 300 K and M7 III 3 214 K. The paper's own M6 and M7 giants average 3 112 and 3 114 K. The
temperature and the correction at it both feed the drawn radius, so the 29 catalogue giants from
M6 to M7.9 were drawn too small: Rho-2 Ari 57.1 solar radii, now 81.5; 30 Her 88.3, now 126.0;
Eps Oct 64.9, now 92.6. The comment said the scale was held "past M7.75"; it now says from M6.
The giant test only reached the third piece (Aldebaran K5 III, Antares M1). It now checks each
piece against table 8: G8 III 4 797.08 K, K2 III 4 387.58, M5.5 III 3 343.43, and M6 and M7 III at
3 134. Controls, each failing "reads a giant's temperature and correction both off its type"
alone: the third piece carried past M6 again; a G giant indexed as K; the first piece's slope
52.74 -> 45; the second's 199.41 -> 180. Before this, the last three left all 817 tests passing.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A non-giant with a spectral type but no colour took its temperature off the textbook colour
spectralTypeToColorIndex gives its type, read on Pecaut & Mamajek's table, and its bolometric
correction off the old textbook anchors. The table puts those colours elsewhere: M8's B-V 1.88
is M5-M5.5 there, 3 001 K where M8 V is 2 570, and the anchors give M8 -3.92 where the table has
-5.65. GJ 3655 (M8, 14.35 pc) was drawn at 0.035 solar radii, a third of Jupiter, where its
type's row gives 0.106 (Mamajek's M8 V: 0.114). Every O type came out B0's 31 400 K.
Both now read dwarfSequenceAtType, which already existed for the giants, and a G magnitude is
carried to V at the type's G-V too. On the published catalogue, of the 835 non-giant stars with
a band, a type and no colour, those more than 200 K off their type's row go from 21 to 0, those
with a radius more than 1.5 times off it from 7 to 1 and a luminosity from 10 to 1 (the one left,
Oph 11, is a G-band star). GJ 3849 (dM9) goes from 0.030 to 0.089 solar radii, GJ 3855 (M6.5)
from 0.040 to 0.087.
Two tests pinned the textbook path and now read the table: M5Ve with no colour is 3 060 K, not
3 106, and K5 tints as B-V 1.15, its row, not 1.105. An O8 star is 35 100 K, its row, not B0's.
Controls: reading the temperature through the textbook colour again fails "is drawn at its
type's row of the dwarf sequence" (with the three updated tests); reading the correction off the
anchors again fails that test alone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The README still said the star was sized against its innermost orbit and made visible by a halo,
two things 7213f98 and 2242fe0 removed. It now says what the view draws: the star at its own
radius on the orbits' scale, the archive's for a host and derived from luminosity and temperature
otherwise, marked "from colour and brightness" on the card; a limb-darkened disc in its
blackbody's colour, lighting its planets in that colour against the Sun's; a grey point where
nothing gives a size or temperature (5aac46d); the three-pixel floor and no halo; the camera held
at 0.05 AU or three radii, whichever is further, and a giant framed inside its neighbours' ring
(d1aa22e).
Measured on :4302 at 1600 by 1000 against published radii: Sun 1.0000 R☉ (card "1.00 solar
radii"); Proxima Centauri 0.1410, the archive's st_rad (0.154 in Boyajian et al. 2012); Sirius A
1.7943, derived, card "~1.8 solar radii, from colour and brightness" (1.711); TRAPPIST-1 0.1192
(0.119). The star's light: Sun (1, 1, 1), Proxima (1, 0.523, 0.165), TRAPPIST-1 (1, 0.443,
0.095), ups And (F8V) (0.899, 0.937, 1), Sirius (0.52, 0.666, 1), intensity pi in every one.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
6d81c45 set ARCHIVE_EPOCH to Gaia DR2's 2015.5 and tested the constant's value, through
propagateProperMotion; nothing tested that resolveHostStarId uses it. The review's mutant that
carried the matcher's queries back from 2016 passed 807 of 807, since the GJ 15 A fixture's decoy
sits 16″ away and half a year moves the query 1.5″ there.
A new case gives resolveHostStarId Barnard's star's archive row (the J2015.5 position and motion
the epoch test already uses) and three stars: Barnard's at the row carried back 15.5 years, and
decoys where carrying it back 16 and 15 years lands, 5.2″ either side at its 10.4″/yr. It must pick
Barnard's. Controls: the matcher carrying back from 2016, or from 2015, each fail it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
d94451e passes the host's surface luminosity, the archive's st_lum where it has one, to the
SystemOrbitsRenderer that classifies each planet, and no test looked at what the renderer got: the
review's mutants that handed it the derived luminosity, none, or the Sun's all passed 807 of 807.
Its commit says 544 planet temperatures move by more than 10 % and 98 planets change class with it.
The scene spec now enters Proxima and checks that Proxima b's marker carries the texture of the
appearance 0.00151 L☉ gives it (228 K, temperate), after checking that null and 1 L☉ each give
another class, so that the texture can tell them apart. Controls: the renderer given
starSurfaceOf(star, []).luminositySolar (the fixture has no measured band, so none), null, or 1
each fail it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
f8af0ec rewrote the neighbourhood note to say where positions come from, and it still said
"measured parallaxes" for everything but the archive's hosts. 313 HYG stars besides the Sun have
neither a Hipparcos nor a Gaia distance and sit at HYG's own, which for these rows is the Gliese
catalogue's resulting parallax: fetchStars says as much ("as often photometric as measured"), and
the review's cross-match with CNS3 (Gliese & Jahreiss 1991, VizieR V/70A; not repeated here)
found about 154 of them with a photometric or spectroscopic parallax, 130 within 25 pc — GJ 3522
drawn at 4.46 pc, 1000/224 mas, with no trigonometric parallax behind it. positionsNote now counts
the HYG stars without a distance error, the Sun aside, and names them. In the app on :4302 the note
reads "Positions from measured parallaxes; for the 313 stars only the Gliese catalogue places, from
its distances, about half of them photometric; for the 3,277 planet hosts only the NASA Exoplanet
Archive places, from its distances." and still fits the readout panel in four lines.
Test: a new star-readouts case counts a Gliese row and leaves the Sun out; the archive-only case
now expects the semicolon. Controls: dropping the Gliese clause, and counting the Sun among the
Gliese stars, each fail it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The card and the body page said "Its host star is not in the catalogue" for every planet without
an equilibrium temperature. Over the shipped data that is 2 714 planets, and the host really is
missing for 27. 2 420 have no semi-major axis, and since 869635b gave a star no survey measured no
luminosity, 267 more have a host that is on the map: OGLE-2005-BLG-390L b's card said so inside its
own host's system. The text now names both things that can be missing without claiming which:
"No temperature could be derived: its star's luminosity or its orbit's size is not known." In the
app on :4302, /body/OGLE-2005-BLG-390L b and /body/PSR J1719-1438 b read it.
Test: the object card's missing-temperature case now expects the new sentence and checks the old
claim is gone. Control: putting the old sentence back fails it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
formatLuminosity printed three decimals from a thousandth to 1 L☉, which leaves one figure below
a hundredth. d94451e put the archive's own luminosity on the card, and Proxima's 1.51×10⁻³ L☉
(st_lum −2.821 ± 0.02 dex) read 0.002, 32 % over and six times the archive's error bar, while
8.9×10⁻⁴ just below the cut kept its two figures. Below 0.01 it is now two significant figures.
Over the 4 440 hosts exoplanets.json gives a luminosity for: 23 printed more than 10 % off it and
43 more than 5 % before (worst 32.5 %); none more than 5 % after (worst 4.4 %). In the app on
:4302, Proxima's card reads "Luminosity 0.0015 L☉".
Tests: quantity.spec now expects 0.0017 → "0.0017 L☉" and Proxima's figure → "0.0015 L☉", and
keeps 0.0523 and 0.523 at three decimals; star-readouts.spec and the scene spec now expect
"0.0015 L☉" where they pinned "0.002". Controls: three decimals below a hundredth, and two figures
reaching up to 1, each fail the named test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
4191b70 carried BP−RP to the dwarf's B−V before the field's tint ramp read it, and said the Gaia
stars had been tinted redder than their own disc in the system view. They had been paler: the ramp
was white at B−V 0.8, a K0 dwarf, where a blackbody against the display's D65 is white near 6 500
K, B−V 0.44, and its red end, (1, 0.6, 0.35), was paler than an M dwarf's disc. After it 245 850 of
the 376 660 BP−RP stars were tinted bluish, 227 797 of them while their disc was warm. The field
now takes blackbodyColor at effectiveTemperatureK, the same function and the same temperature the
disc is drawn with, so a giant is tinted at its type's temperature too.
Over the shipped catalogue: stars bluish in the field with a warm disc 227 797 → 0, bluish at all
245 850 → 17 931, mean RGB distance from each star's field tint to its disc 0.250 → 0.001 (what is
left is the tint being cached at the nearest 10 K, which moves no channel by more than 0.0014). In
the app on :4302, field against disc after entering each star: Gaia DR3 6361559602963567744 (BP−RP
0.882, 5 543 K) (0.974, 0.982, 1) → (1, 0.854, 0.764) against (1, 0.854, 0.765); Gaia DR3
2026408043220034176 (1.84, 3 850 K) (1, 0.793, 0.664) → (1, 0.629, 0.34), the disc's own; HD 13531
(G0, B−V 0.70) (0.971, 0.979, 1) → (1, 0.86, 0.779), its disc's.
A blackbody per star cost 100 ms, and the scan of the table rows another 70, so the tint is cached
per 10 K and the table is bisected. Building the star field over the whole catalogue in the app
took 176-196 ms before and 151-159 after, four runs each.
Tests: a new case tints a G2 dwarf warm, exactly its 5 770 K blackbody, and Antares at the
temperature its type gives; the BP−RP render test now also checks an M0's blackbody. Controls: a
giant tinted at its colour's temperature, the tint taken against a bluer white than the disc's,
and a cache a hundred times too coarse each fail the named test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
d097f4b gave a giant its type's bolometric correction but left its temperature at the dwarf its
colour reads as, and the radius drawn from the two paired a correction for one star with the
temperature of another. The M giants stayed too cool (610 of them at a median 3 275 K, Antares
3 019 against Ohnaka et al.'s 3 660), and a hot giant behind dust got a 30 000 K star's correction
at the temperature of an A star: Menkib, O7.5 Iab at B−V 0.02, was drawn at 9 517 K and 95 R☉,
Alp Cam, O9.5 Ia, at 338, where 14 and 21 are published. Rigel went from 81 to 102 R☉, Alnilam
from 57 to 108.
giantSurface now reads both off the type: G to M giants off van Belle et al.'s (2021, table 8)
interferometric scale, fitted to 191 giants from G1 to M7.75 III, with the correction the dwarf
sequence has at that temperature; O to F giants off the dwarf of their type, for which the table
gains Mamajek's O3 to O9.5 rows (without colours, which do not tell O types apart); carbon and S
stars, which no row reads and which got the Sun's −0.06 at 2 420 K, off the medians of Bergeat et
al. (2001): 2 990 K over the 441 stars of their table 10 and −2.83 over the 383 with a V magnitude,
counted again from VizieR here.
On the shipped catalogue (drawn radius in R☉, before → after, published): Antares 690 → 410 at
3 730 K (680; its luminosity from V is 0.4 dex under Ohnaka's), Aldebaran 48.5 → 44.0 (44.2),
Arcturus 22.5 → 24.1 (25.4), Menkar 160 → 103, Gacrux 118 → 73, Rigel 102 → 67 (74.1), Alnilam
108 → 33, Menkib 95 → 6.9 (14, the dust still dims it), Alp Cam 338 → 31, La Superba 133 → 311
(315) and 544 → 6 977 L☉ (8 090 from Bergeat's bolometric magnitude), 19 Psc 130 → 305 (295).
The 610 M giants now sit at a median 3 644 K (p10 3 386, p90 3 816). Against their own radii
before, the O giants' fall to a median 0.08, the B giants' to 0.62, the M giants' to 0.68, and the
K giants' rise by 8 %. It is not better everywhere: Pollux goes from 8.6 to 10.1 against 8.8,
119 Tau from 700 to 326 against 587, and Mintaka and Alnitak, placed by Hipparcos at 212 and 226
pc where they are about 380, come out 8.8 and 11.8 against 13 to 20.
Tests: the giant case in stellar.spec now checks Antares's temperature against Ohnaka's, and
Aldebaran (to a tenth) and Rigel (to a fifth) against their interferometric radii; new cases give
Menkib its type's 36 100 K and a radius within 2.5 times the published one, and La Superba a
luminosity within a fifth of Bergeat's and 2 990 K; spectral.spec covers dwarfSequenceAtType. The
scene's supergiant case now expects Antares at 350-480 R☉ where it pinned 600-760. Controls:
the luminosity or the temperature ignoring giantSurface, G-M giants read as the dwarf of their
type, their correction taken off the type instead of their temperature, O giants through the
textbook colour clamped at B0, the type index off by one subclass and carbon stars unhandled each
fail the named test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
31e0c04 clamped an untyped star bluer than BP−RP −0.12 to the table's B9 row, so every one of the
110 white dwarfs within 50 pc was drawn at 10 700 K. Gentile Fusillo et al. (2021, MNRAS 508, 3877)
fit 104 of them at 14 266 to 39 304 K, and the radius their mass and gravity give was a median
0.65 of the one drawn. Past the table's end, dwarfSequenceAtColor now reads BP−RP off the median
pure-hydrogen temperature they fit in bins of ±0.025 around −0.15 to −0.40 (15 369 to 28 585 K,
counted again from the cross-match: 27, 24, 22, 15, 10 and 2 stars), and the correction and G−V
off the table's own rows at that temperature, through a new dwarfSequenceAtTemperature that
temperatureToColorIndex now shares.
Against GF21's R = sqrt(GM/g) over the same 104, the drawn radius goes from a median 1.53 (p10
1.33, p90 1.96) to 0.96 (0.94, 1.05), and the temperature from 0.59 of theirs to 1.00 (0.88,
1.02). Gaia DR3 6791196382856581376 is now 19 251 K and 0.0120 R☉, against their 19 205 K and
0.01245.
Tests: stellar.spec's −0.25 case now expects 19 012 K where it pinned 10 700, and a new case gives
that white dwarf its radius to within a fifth; spectral.spec reads −0.15, −0.13 and −0.6, keeps
the red end and B−V's blue end at their rows, and covers dwarfSequenceAtTemperature. Controls:
no white-dwarf branch, the bin's temperature without interpolating, B9's correction kept at the
new temperature, and the temperature read the wrong way round each fail the named test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The search index and the route index gave every one of the 455 571 stars a subtitle up front
through spectralClassification, which filtered the dwarf table afresh on each call; the star
field's tints read the same table once per BP−RP star. Both indices now carry the star itself
and classify it only for the rows shown (entrySubtitle): eight search results, a few route
options. The two filtered columns of the table are built once.
Node, over the shipped catalogue, five runs: classifying every star 121-167 ms before, 56-62
after hoisting the table alone; reading the sequence at every BP−RP colour 187-216 ms, now
111-127. In the app on :4302, two cold loads each, the long task when the Search tab opens was
264-380 ms (median 280) before and 164-276 ms (median 171) after; the longest boot task 863 and
875 ms before, 654 and 741 after.
Tests: the search spec checks its index holds no classification and the row still reads
"Star · ~M8"; the scene spec now reads the route options, and checks its index holds none either.
Controls: classifying every star in the search index, doing so in the route index, and route
options printing the index's subtitle each fail the named test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
c64eea0 left out 41 naked-eye stars "because neither survey gives them a distance". Hipparcos does:
HYG's distances are 1 000 over van Leeuwen's 2007 parallaxes, which the ETL already downloads for
their errors, but HYG writes its 100 000 pc placeholder for every parallax under 1 mas whatever its
error, while keeping far less certain ones above it (Alnilam at 1.65 ± 0.45 mas is drawn). 30 of
the 41 have a positive parallax in the new reduction and 7 have one at least 2.5 times its error:
HD 74180 at 0.67 ± 0.16 mas (4.2 σ), Mu Cep at 0.55 ± 0.20.
hipparcosDistancePc (star-merge.ts, with HYG's placeholder constant moved beside it) keeps HYG's
distance and, where it gives the placeholder, takes 1 000 over the parallax when that is 2.5 times
its error; fetchHipparcosParallaxErrors now returns the parallax with its error. From cache the ETL
keeps 73 563 HYG rows against 73 556 and publishes 455 571 stars; the seven are Alp Cam (1.4 to 3.0
kpc), Psi-1 Aur, HD 74180 (1.2 to 2.0 kpc), HD 86352, HD 96918, Mu Cep (1.3 to 2.9 kpc) and HD
217476, each printed as the range its parallax's error gives. The other 34 stay out: no parallax,
or one within 2.5 times its error of zero. Parallax distances printed as ranges go from 1 109 to
1 116; the validators hold.
Controls, each failing its named test: placeholder rows given no distance, and any parallax 1.25
times its error placing a star (1 of 807 each).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
placementDistancePc took Gaia's distance wherever the cross-match gave a usable one, and its
docstring and c64eea0 called that "the better measurement", adding that Gaia saturates on the
brightest stars "so those sit at their Hipparcos distance". For 273 of the 88 781 HYG stars with
both, Gaia's relative error is the larger, 257 of them naked-eye and 38 by more than twice: Eta Leo
was drawn at Gaia's 556.6 pc ±17 % against Hipparcos's 389 pc ±6.2 %, Schedar and Tarazed at
Gaia's though both have Gaia G of about 2.
placementDistancePc now takes both relative errors and picks the smaller, Gaia's as before when
either is missing; fetchStars gives the placed star the error and the Gaia-distance flag of the
distance it took. The merge's combine keeps the Gaia entry's direction, and now takes the other
entry's distance with its error where that is the more precise, so a bright star Gaia's main query
also holds lands at its Hipparcos distance too. Gaia still wins for the rest, whose parallaxes are
some fifty times more precise.
From cache: 424 stars move, 386 of them naked-eye; HYG rows at Gaia's distance 58 379 -> 58 109 and
HYG stars flagged so 8 129 -> 8 095. Eta Leo 556.6 -> 389.1 pc, Tarazed 178.9 -> 121.1, Imai 139.5 ->
105.8, Schedar 71.0 -> 70.0, Alphecca 23.7 -> 23.0, each now "±" its Hipparcos error. Where the two
disagree by more than 5 σ (21 stars, Iot Gem 11 σ), one of the formal errors is wrong, and this
takes the smaller one at its word. The validators pass unchanged.
Controls, each failing its named test: the placement ignoring the errors, the merge keeping Gaia's
distance whatever the errors, and taking the distance but not its error (1 of 805 each).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
For an archive-placed host with a temperature and no B magnitude, fetchExoplanets took B−V from
Ballesteros' blackbody fit, which runs 0.1 to 0.2 redder than Pecaut & Mamajek's dwarf sequence
below 3 800 K, and the luminosity then read its correction off that sequence at that colour: 3 500 K
came back as 3 102 K with a correction 1.15 magnitudes too large, anything under about 3 170 K was
clamped to B−V 2.00, and the card printed it as a measured "Colour B−V 2.00". CFBDSIR
J145829+101343, a 580 K brown dwarf, read "Spectral type ~M6, from colour".
temperatureToColorIndex now reads the table itself backwards, interpolating B−V between the two
types the temperature falls between, so the temperature and correction read back off the colour
are the table's at that temperature; it has no answer outside 2 420 to 31 400 K. The colour is
flagged colorFromTemperature, a fifth bit in the photometry byte (the format, README and the ETL's
round-trip check follow), and the card prints it "B−V 1.66, from its temperature", marked derived.
From cache: 57 archive stars change colour; 54 carry the flag and 3, CFBDSIR J145829+101343 among
them, now have none. For the 47 of them the archive gives a luminosity, the one derived from
magnitude and colour moves from a median 0.228 dex off it to 0.124; Kepler-445 (3 157 K) from
0.0282 L☉ to 0.0080 against the archive's 0.0079. Its colour goes from 2.00 to 1.67. The card shows
the archive's luminosity where it has one since earlier on this branch, so this is the figure used
for the rest and for their radii.
Controls, each failing its named test: the nearest hotter row taken without interpolating (2 of
803 failed), no refusal outside the table, the flag not encoded, and the card calling the colour
measured (1 of 803 each).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Three things the branch added could vanish from the data with every check passing, as the review
showed with guarded ETL mutants: dropping st_rad or st_teff from each planet left 0 of 6 354
carrying their host's radius or temperature, storing Gaia's parallax_error in milliarcseconds
rather than over the parallax changed 431 464 distance readouts, and dropping the flag fetchStars
sets relabelled 8 129 HYG stars placed at Gaia's distance "HYG". The round trip compares decoded
with encoded, and the counts of missing bands and errors do not look at what the values are.
validateExoplanets now requires 90 % of planets to carry their host's radius and temperature
(measured 6 030 and 6 054 of 6 354). validateStars requires the median relative error of the
distances at Gaia's to be under 1 % (measured 0.33 %), at most 1 % of stars to have a parallax
distance with an error of a fifth or more (measured 1 109, 0.24 %; the archive's, which are on the
distance and never ranged, are left out), and at least 6 000 HYG stars flagged at Gaia's distance
(measured 8 129). Figures from an ETL run from cache on this branch.
Each proved by a guarded mutant in a scratch copy of the ETL, run from the cache, failing with its
own message: host radius dropped ("Only 0 of 6354 exoplanets carry their host's radius and 6054 its
temperature"), host temperature dropped ("6030 ... and 0"), the Gaia error left in mas ("The median
error of Gaia's distances is 1.92 %"), the same with the median check waived ("17689 parallax
distances have an error of a fifth or more"), and the flag dropped ("Only 0 HYG stars are flagged
at Gaia's distance"); a no-op edit completed. No ceiling is loosened.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
fetchExoplanets placed each star it adds from the Exoplanet Archive by carrying the archive's
position back sixteen years, and 4c8e4a0 said the archive publishes at Gaia's J2016. It does not:
its positions are Gaia DR2's at J2015.5 although it names the DR3 source. Barnard's star in the
archive, 269.4486144, 4.7379808, equals DR2 4472832130942575872 at J2015.5 to 1e-7°, and sits 5.2″
from DR3's J2016 place; the review found 66 of the 67 archive-placed stars moving over 100 mas a
year within 0.21 mas of their DR2 position and none within 1 mas of DR3's. The matcher's comment
made the same claim of HD 133131 and TOI-2459, whose positions are DR2's too.
host-star-matching.ts now exports CATALOGUE_EPOCH and ARCHIVE_EPOCH = 2015.5, the matcher carries a
query back by their difference, and fetchExoplanets uses the same two, so the epoch is written
once. From cache: only the 3 277 archive-placed stars move, 2 287 of them by more than a
milliarcsecond, 74 by more than 50 and TOI-2406 most, by 203 mas, half a year of its 405 mas/yr;
no planet changes host. Invisible at map scale; the point is that the constant and its comments
now say what the archive does.
Control: ARCHIVE_EPOCH back to 2016.0 fails "carries Barnard's star from the archive's position to
where Gaia DR3's goes, to a few milliarcseconds" (1 of 802), which puts the two 5.2″ apart. The
GJ 15 A matching fixture is now built at J2015.5 too. fetchExoplanets' use of the constant has no
test of its own; the ETL has no test harness.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
4c8e4a0 numbered each star it adds from the Exoplanet Archive by its place among the unmatched
hosts, in pl_name order, and the refresh workflow re-queries the archive every Monday. One host
added or dropped ahead of another renumbers it: in the review, removing a single planet row
renamed 3 276 of the 3 277 ids, and a bookmark kept on Kepler-186 (1070001620) opened Kepler-1860
under Kepler-186's stored name. HYG's and Gaia's ids do not move between refreshes; the merge's own
comment says ids are meant to hold.
archiveStarId (host-star-matching.ts, beside the matcher the ETL already imports from there) hashes
the host name with FNV-1a into the 3.7 million ids between 1 070 000 000 and 2^30, and moves a name
whose id is taken to the next free one. The added stars are sorted by id before they are appended,
so the published list stays in id order. Measured: all 4 237 archive-hosted planets change host id
once (Kepler-186 is now 1073671518), none of the others; one of the 3 277 names was probed past a
collision, and one new id falls in the old 1070000000-1070003276 range, so a bookmark saved on that
old id would open the wrong star once. Dropping the same row from a copy of the cached answer and
rerunning the exoplanet step in a scratch copy of the ETL now leaves 3 276 of 3 276 ids unchanged.
The validators pass: no duplicate id, all under 2^30.
Controls, each failing its named test: the id taken from the order of arrival, and no probing past
a taken id (1 of 801 each).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
c01d3ec read OpenNGC's addendum, which brought in its only two DrkN rows, C099 the Coalsack and
B033 the Horsehead; NGC.csv has none. classifyOpenNgcType mapped DrkN to 'nebula', so both were
drawn as the backdrop's nebula sprite: ff86b0, additive, at the faintest opacity and the 70 pc
minimum size, about 1.6° across. The review measured it adding up to +59 in red over the Coalsack
beside Crux, where the sky has a dark hole, and darkening no pixel, as an additive sprite cannot.
DrkN is now among the codes dropped, with the reason beside the map. fetchDeepSky from cache keeps
488 objects against 490, still 107 Messier objects and every required id; README's count follows.
On :4302 the served backdrop holds 488 objects and the scene 488 sprites, neither of them C099 or
B033.
Control: DrkN mapped back to 'nebula' fails "leaves out a dark nebula, which a sprite that adds
light cannot draw" (1 of 801).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
74b93a0 let a Gliese-only HYG row fold into a Gaia entry that moves with it up to a minute of arc
away, since Gliese's positions are off by that much. Its report counted only the 3-60″ residue.
Measured on the published catalogue with the merge's own motion and brightness rules, 39 Gliese
rows within 25 pc still had a bare Gaia entry moving with them 60 to 150″ away (32 within 100″,
none past 150); with every row shifted a quarter of a degree north or south, none did. The review
found 36 of them to be the same star under SIMBAD, three within 10 pc: GJ 3618 (LHS 288) at 4.49 pc
beside its Gaia entry at 4.83 pc 94″ away, GJ 1123 at 8.13 and 9.52, GJ 1001 at 9.60 and 12.31.
MERGE_COMOVING_ANGULAR_TOLERANCE_DEG is now 160″. The ETL from cache folds 62 046 entries against
62 002, leaves 11 510 HYG rows without a counterpart against 11 554, and publishes 455 564 stars;
within 10, 25 and 50 pc the map now holds 369, 5 523 and 40 877 against 372, 5 564 and 40 921
(GCNS: 312 Gaia sources within 10 pc). No Gliese row within 25 pc has a co-moving bare Gaia entry
3-300″ away any more. GJ 3618, GJ 1123, Gl 319C and GJ 3999A now sit on Gaia entries. One
exoplanet changes host id: LHS 475 b's Gaia entry took HYG's id with its description.
Not every name lands where SIMBAD would put it: Gl 319C takes the entry SIMBAD calls GJ 319 B, as
HYG's Gl 319B already sat on SIMBAD's C, and GJ 3999A takes the entry named GJ 4000B before, which
moves to the bare one beside it. Both systems go from one entry too many to one per star; the
names follow HYG's positions, which only identifiers could correct. The merge validators hold: 23
cross-catalogue pairs within an arcsecond, as before.
Controls, each failing its named test: the window back to a minute (2 of 801 failed), and 200″
(1 of 801).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The comment on hostStarRadiusSolar and its two neighbours said they were read from "the planet's
default row first, then the composite table". The default-row query (TAP_COLUMNS in
fetchExoplanets.ts) asks for none of st_rad, st_teff or st_lum — the cached answer's columns end at
st_mass and disc_year — so host() always reads them from pscomppars, whose columns may each cite a
different reference: Proxima's 0.141 R☉ is the composite table's. It also said they were measured
where stellar.ts derives a luminosity, which held for the luminosity only once starSurfaceOf began
reading st_lum, earlier on this branch; it now points at starSurfaceOf. A comment, so there is no
test to fail.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
a81dd49 added a query of public.hipparcos_newreduction on the ESA archive, cached as
hipparcos-errors-<hash>.csv, and fetchStars requires it: a failed or short answer throws. The
refresh workflow carries only tools/etl/.cache/gaia-dr3-*.csv between runs, a glob that file does
not match, so every weekly run fetched its 117 955 rows live from the archive the cache exists to
spare, and an ESA outage would have failed the refresh even with every Gaia answer cached. Before
a81dd49 a warm cache meant no request to that archive at all.
The cache step now lists both globs. Its key gains "-hipparcos": actions/cache keys are immutable,
and an entry already saved under the old key would be restored without the new file and never
saved again. Checked with Node's path.matchesGlob against the local cache (gaia-dr3-54fdbc7a,
gaia-dr3-hip-785b92fc and hipparcos-errors-f846b045 match, the archive's own files do not) and by
parsing the workflow with PyYAML. A workflow file has no unit test to fail without the change.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
describingCatalogue calls a star Gaia's when its source is gaia and its band is not V. 44 Gaia
sources in the shipped catalogue have no G (Gaia DR3 40091256260676736 among them), and nothing
tested them: narrowing the rule to band G, which would relabel all 44 "HYG" and count them as HYG
in the neighbourhood census, passed the suite. The existing test already builds such a star and
now checks its Source row as well.
Control: the rule narrowed to band G fails "says a stand-in magnitude was not measured, and leaves
out a colour there is none of" (1 of 797).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
5333be6 gave the star card an estimated type for the 383 695 stars the ETL files as "Unknown",
but search rows, the route search index and the current-star route option still printed that
literal: TRAPPIST-1 read "STAR · UNKNOWN" in search beside "Spectral type ~M8, from colour" on its
card, audit #17's own symptom. spectralClassification (spectral.ts) now gives the catalogue's type,
else the colour's marked "~", else nothing, and all three and the card's subtitle use it; a search
row with nothing to add prints its kind alone rather than a trailing " · ".
The neighbourhood note still read "Positions from measured parallaxes" after 4c8e4a0 placed 3 277
stars at the Exoplanet Archive's sy_dist, 343 of them without a usable parallax and 281 of those
past 1 kpc — microlensing hosts such as OGLE-2005-BLG-390L at 6.6 kpc, from a lensing model.
positionsNote now says so, with the count. Neither the note nor the census subtitle (audit #16) had
a test: putting back "Hipparcos · Yale Bright Star · Gliese" passed all 775.
In the app on :4302: the note reads "Positions from measured parallaxes, and for the 3,277 planet
hosts only the NASA Exoplanet Archive places, from its distances. Grid marks the galactic plane
through the Sun."; search reads "TRAPPIST-1 STAR · ~M8", "OGLE-2005-BLG-390L STAR", "Sirius STAR ·
A0M"; TRAPPIST-1's route option has subtitle "~M8".
The scene spec gains an archive-placed fixture star. Controls, each failing its named test: the
search row printing the raw type, keeping the separator with nothing after it, the route index
and the current-star option printing the raw type, the note back to parallaxes only, the subtitle
back to the old catalogue list (1 of 797 each), the note never counting the archive (2 of 797),
and the classification without its estimate (3 of 797).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
colorIndexToRgb maps a B−V onto the star field's tint ramp, and its one caller passed it every
star's colour index, whatever colorSystem said. BP−RP is the larger of the two for the same star —
0.823 against 0.65 for a G2 dwarf, 1.84 against 1.42 for an M0 (Pecaut & Mamajek) — so the 376 703
stars measured in it, 83 % of the catalogue, were tinted as later types than they are, and redder
than HYG's stars of the same type, and than their own disc in the system view.
A BP−RP is now carried to the B−V of the dwarf sequence at that colour (the table's own B−V
column, which DwarfSequencePoint now returns, clamped at its ends), then tinted as before. The
median Gaia colour, BP−RP 0.883, a G7 dwarf, read as a K2: its tint goes from (1, 0.972, 0.955) to
(0.975, 0.982, 1), the same as HYG's G7 stars.
Controls, each failing its named test: BP−RP read as B−V (2 of 792 failed), the star field not
passing the colour system (1 of 792), and the B−V taken from the wrong column (3 of 792).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
formatDistance printed any error of a fifth or more as the range a parallax's error makes,
d/(1+e) to d/(1−e). A star the ETL places from the Exoplanet Archive carries the mean of sy_disterr1
and 2 instead, which the archive defines as one-sided errors on sy_dist in parsecs, and many of
those distances come from a lensing model with no parallax behind them. 129 of the 3 195 archive
stars with an error were printed as ranges the archive does not give: KMT-2016-BLG-1836L as 5.8 to
9.2 kpc, where the archive publishes 7 100 +800 −2 400 pc, and AT2021ueyL as 664 pc to 2.4 kpc
against 1 040 +740 −440.
The card now tells formatDistance when the error is on the distance (source exoplanet-archive), and
it keeps the ± at any size. Measured on the shipped catalogue: KMT-2016-BLG-1836L reads 7.1 ± 1.6 kpc,
AT2021ueyL 1.0 ± 0.6 kpc, EPIC 201170410 134 ± 43 pc, KMT-2016-BLG-0212L 6.3 ± 1.3 kpc. The ETL keeps
only the mean, so a lopsided interval such as KMT-2016-BLG-1836L's is shown symmetric; carrying both
errors would take a second column in stars-meta.bin. The doc comments on formatDistance and on
distanceError no longer say an archive distance is an inverted parallax.
Controls, each failing its named test: the error on the distance read as a parallax's (2 of 790
failed), and the card not saying it is on the distance (1 of 790).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The system view names a star's neighbours on a ring at 0.74 of the view's tighter half-extent,
whatever the star. A giant was framed to fill the frame, 2.4 radii out, which the three-radius
closest approach then overrode, so it settled at 3 radii with its disc projecting to 0.758 of the
half-extent: in the app, Betelgeuse's disc was 379 px on a 1 000 px view against the 370 px ring,
and HD 39374 and HD 38118 were printed on it at about 1.3:1 contrast.
The star's term in the framing distance now places its silhouette, asin(R / d), at half the
tighter half-extent, which puts the camera 4.4 radii out on a 50° view. Measured on :4302 at
1600 by 1000: Betelgeuse settles at 13.9 AU (closest approach 9.5) with a 250 px disc and Antares at
14.1 AU; the nearest corner of any neighbour's name is 292 px from the centre, so none is on
either disc (it was two and one of them before). Closing in to the closest approach can still bring
the disc over the names; that is the viewer's choice, not the arrival's.
This also gives the star's term a job: at 2.4 radii it never exceeded the closest approach, so
dropping it changed nothing (the previous commit's one surviving control). Controls, each failing
its named test: the star left out of the framing (1 of 788 failed), the star framed to fill the
frame (2 of 788), and the disc taken as flat (1 of 788).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The system view drew every star without a radius at the Sun's radius, and every star without a
temperature in the Sun's colour. PSR J1719-1438, a neutron star 10 km across, came out 1 R☉ wide,
wider than its planet's 0.0044 AU orbit, and b spent nine tenths of its orbit inside it; Procyon B,
a white dwarf of 0.012 R☉ with a type the parser does not read and no colour, was a Sun 81 times
too wide in the Sun's colour.
Such a star is now drawn at 150 km, under the three-pixel floor from anywhere the camera can go, so
the floor's point is what shows, and its disc keeps the photograph's own grey instead of the Sun's
tint. Its light stays white, which is the photographs' own light rather than a claim about the
star. 2 858 stars take this after the clamping in the previous commit, against 3 077 drawn at the
Sun's radius before it.
In the app on :4302: PSR J1719-1438's star is drawn at 1.7×10⁻⁴ AU (the floor, 168 times its
geometry) with b 0.00417 AU from its centre, outside it; Gl 280B at the floor, tint (1, 1, 1) and no
radius on its card; Proxima at 0.141 R☉, tint (1, 0.459, 0.135), light (1, 0.523, 0.165), card
"Luminosity 0.002 L☉" unmarked and "Radius 0.141 solar radii".
The scene's use of the star's surface had no test: every star in the scene spec was drawn at the
Sun's radius, and drawing all of them so, lighting planets white, tinting every disc as the Sun or
closing the camera to the Sun's clearance all passed. The spec now gives Proxima its planet's
archive row and adds Antares and Procyon B, and three tests enter them. Controls, each failing its
named test: every star at the Sun's radius (3 of 787 failed), an unmeasured one at it, the light
without the temperature, the closest approach for the Sun, the Sun's tint for a host, the Sun's tint
for a star without a temperature, and the card without the surface (1 of 787 each). Dropping the
star from the framing distance was not caught: its term, 2.4 radii, never exceeds the three-radius
closest approach the camera is clamped to, so it cannot change where the camera settles.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
dwarfSequenceAtColor answers null outside Pecaut & Mamajek's table, B−V −0.301 to 2.16 and BP−RP
−0.12 to 5.1, and effectiveTemperatureK then fell back on the type, which Gaia's stars do not have
and carbon stars' parser does not read. 219 stars with a measured colour got no temperature and so
no radius, and were drawn at the Sun's radius in the Sun's colour: 110 white dwarfs within 50 pc,
38 Gaia stars redder than BP−RP 5.1 (Gaia DR3 6439125097427143808, an ultracool dwarf 4.0 pc away),
HD 46687, La Superba and the other carbon stars, and an O8 star.
Past the table, the colour is now read at the row it is past (clampToTable), but only for a star
with no readable type: beside a type an off-table colour is more often the bad measurement — HD
49748 is G5 V at B−V −0.32 — so the type still wins there, as it did. The luminosity reads the same
point, so a star past the red end also gets the M8.5 row's G−V and correction. A type's own colour
past the table, which only O types have, is read at B0. Carbon and S stars (C, N, R, S) now count
as giants, so their correction stays their type's, not an M8.5 dwarf's −5.78.
Measured on the shipped catalogue: stars without a temperature 3 053 -> 2 834, without a radius
3 077 -> 2 858; all 292 stars with an off-table colour now have a temperature, against 73. Gaia DR3
6439125097427143808 is 2 420 K and 0.110 R☉ (M8.5 V: 0.104); the 110 white dwarfs a median 0.018 R☉
at 10 700 K (0.0013 to 0.034); HD 46687 212 R☉ at 2 420 K and La Superba 133; audit #16's Gaia DR3
5612323414549657984, k1 Pup, B6 V at BP−RP −0.15, 203 L☉ and 4.1 R☉ against 143 and none (FLAME
gives 336 and 3.49). The 2 851 stars left without a radius have neither a colour nor a readable
type, or no band.
Controls, each failing its named test: no clamp for an untyped star (2 of 784 failed), the clamp
winning over a type, an O type's colour unclamped, the luminosity off the unclamped sequence, and
carbon stars not counted as giants (1 of 784 each); clampToTable ignored (3 of 784).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
3392f06 read every star's bolometric correction off the dwarf sequence at its colour, so a giant
got the correction of the cooler dwarf of that colour, which is larger. Antares, M1 Ib at B−V 1.87,
took an M5 dwarf's −3.26 and came out 172 023 L☉ and 1 516 R☉, a 7.06 AU sphere, against the 680 R☉
Ohnaka et al. (2013) measure; 119 Tau 2 838 against 587, Menkar 204 against 89.
isGiant (spectral.ts) reads luminosity class I to III off a type's primary component, or HYG's g
and c prefixes, and a giant keeps its type's correction. Drawn radius against the published one
(no planets, so derived), before and after: Antares 2.23 -> 1.01, 119 Tau 4.84 -> 1.43, Menkar 2.29
-> 1.80, Scheat 1.92 -> 1.42, Aldebaran 1.95 -> 1.10, Mirach 1.96 -> 1.27, 41 Com 2.14 -> 1.56;
Betelgeuse 0.76 -> 0.89. It is not better everywhere: Gacrux goes from 1.23 to 1.40 and Arcturus
from 1.02 to 0.88. 10 808 stars have a giant's type, 10 794 of them a colour. Against the archive's
own luminosity for the 130 giant hosts, the median error moves from 0.038 to 0.052 dex and the 90th
percentile from 0.246 to 0.204; 17 are off by over 1.5 times, against 19.
Controls, each failing its named test (1 of 781): the giant given the dwarf's correction, a giant
companion read as the primary, a IV read as a I, and the prefixes ignored.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The ETL has carried each host's st_lum since 494fb56 and nothing read it: the card, the system
view's planet temperatures and the body pages all derived a luminosity from the star's magnitude
and colour. For 757 of the 4 440 hosts the archive gives one for, the two differ by more than
1.5 times, 667 of them stars the ETL placed from the archive's own V and B−V. Proxima read
8.9×10⁻⁴ L☉ beside the archive's radius and temperature, which imply 1.27×10⁻³; the archive gives
1.51×10⁻³, as Ribas et al. 2017 do.
starSurfaceOf now returns the luminosity with the radius and temperature, taken from the first of
the host's planets that gives one, as st_rad and st_teff already are, and derived only otherwise.
The system view lights its planets with it, the body page uses the same host rule, and the card
marks it derived only when it is. The derived radius still comes from the derived luminosity, so
its "from colour and brightness" stays true; only 3 hosts have st_lum and no st_rad.
Measured on the shipped catalogue against the parent commit: 544 of 3 639 planets with an
equilibrium temperature move by more than 10 %, and 98 change class. Kepler-186 f goes from a 292 K
rocky world to a 175 K icy one, LHS 1140 b from 152 K icy to 206 K temperate, LTT 1445 A b from
298 K rocky to 402 K scorched, Proxima Cen b from 200 to 228 K and d from 259 to 296 K.
starReadouts now takes the surface alone, which carries the luminosity. Controls, each failing its
named test: the archive's luminosity ignored (2 of 778 failed), called derived, the body page
warmed by the derived one, the body page reading the planet's own row only, and the card marking
a published luminosity derived (1 of 778 each).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
luminosityOf derived a luminosity from any magnitude, including the stand-in the ETL writes for
the 309 stars with neither a V nor a G. starSurfaceOf already refused it for the radius; the card,
the system view's planet appearances and the body pages did not. The 265 archive hosts among them
came out at a median 33 L☉: KMT-2016-BLG-1107L, a 0.087 M☉ star, at 36.8, and OGLE-2005-BLG-390L b,
published at about 50 K, read 388 K and a sub-Neptune. Their cards printed "Magnitude: Not
measured" beside a luminosity derived from that magnitude.
The guard now sits in luminosityOf, which every caller goes through, so the one in starSurfaceOf
is gone. The Sun keeps its 1 whatever band it is filed under. Measured on the shipped catalogue:
of the 275 planets on those 265 hosts, 268 had an equilibrium temperature and 0 now do; before
the archive stars were added they had no host, and so none either.
Controls: without the guard, "has none from a magnitude no survey measured, and gives its planets
no temperature from it" fails (with the radius test beside it, 2 of 777); without the Sun's
exemption, "is 1 for the Sun, whatever its magnitude is filed under" fails (1 of 777).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every star but the Sun was a flat disc of one colour, and the Sun wore the texture pack's orange
photograph, lit by the same white light as every other star's planets.
Every star now shares one TSL material (starSurfaceMaterial in the scene): the Sun's map in grey,
times the colour of a blackbody at the star's temperature, times a linear limb-darkening law,
1 - 0.6 (1 - mu), the Sun's coefficient in the visible. The temperature is the one the radius
uses: the archive's st_teff for a host, else the dwarf sequence at its colour or type, else
the Sun's.
The colour comes from blackbodyColor (stellar.ts): Kim et al.'s cubic fit to the Planckian locus,
then CIE XYZ to linear sRGB, brightest channel 1. At D65 it gives 2 900 K (255, 180, 103), 5 800 K
(255, 241, 235) and 9 600 K (208, 219, 255), against (255, 182, 98), (255, 241, 231) and
(211, 221, 255) in Charity's integrated blackbody table. The tint is a uniform, so the shader is
built once and not per system.
The star's PointLight takes the same colour against the Sun's, since the planets' photographs
were taken in sunlight: the Sun's light stays white at pi, TRAPPIST-1's (2 566 K) is
(1, 0.44, 0.10) and Proxima's (2 900 K) (1, 0.52, 0.17), Sirius's (0.52, 0.67, 1). The intensity
stays pi. No halo comes back.
sun.jpg was 2048 by 1024 and 822 427 bytes for a disc that reaches 216 px across at the Sun's
closest approach on a 1080-line screen. It is now 1024 by 512 in grey, 31 306 bytes, which covers
the disc to a 1440-line screen. Its brightness varied by 56 % rms, which made every star a mottled
rock; it is rescaled to 14 % rms about the display's white, of the order of the Sun's granulation
contrast, the brighter half clipped as in a photograph exposed for the disc (6 % rms remains).
Measured on the dev server (1600 by 1000, the camera at its closest approach):
- the disc's brightness against the law, from r/R 0.52 to 0.97: Sun 0.970/0.841/0.763/0.657/0.560
against 0.927/0.829/0.755/0.645/0.554, ups And and Sirius the same to within 0.05;
- the disc's centre in sRGB: Sun (245, 233, 226), ups And (F8V, 6 157 K) (249, 239, 239),
Sirius (199, 209, 243);
- the first system entry of a fresh page, three runs each in alternating blocks against the
previous commit: Sol's longest task 90 ms before, 86 ms after, two tasks over 50 ms in every
run either way; Proxima Centauri's none over 50 ms after, one run of six with a 53 ms task
before. The long tasks on Sol's first entry predate this change.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The system view drew the Sun at its own radius and every other star at 0.45 of its innermost
orbit, capped at 0.2 AU: a size chosen so the star would not swallow its planets, not the star's.
Proxima Centauri was drawn at 2.8 solar radii, eighteen times its own, and every star without
planets at 43.
starSurfaceOf (body-view-model.ts) now gives each star a radius and a temperature. A planet host
takes the archive's st_rad and st_teff from its planets' rows: 4 439 hosts are drawn at a
measured radius, 22 at a derived one. Every other star's is derived: its temperature off Pecaut & Mamajek's dwarf
sequence at its colour (the same table the spectral estimate reads, or at the colour its type
implies where it has none), its luminosity from its absolute magnitude and the bolometric
correction luminositySolar already applies, and R = sqrt(L) / (T / 5772 K)^2. Against the
archive's own st_rad for the 1 447 catalogue hosts that have one, the derived radius is within
0.018 dex at the median, 0.071 dex at the 90th percentile, and within a factor of 1.5 for
97.1 %. Sirius comes out 1.79 solar radii (1.711 published, Liebert et al. 2005), Wolf 359 0.117,
Betelgeuse 584, the Sun exactly 1.
A star with no band has only the ETL's stand-in magnitude, and gets no derived radius: PSR
J1719-1438 came out 2.3 solar radii from it, wider than its planet's orbit. With the stars that
have neither a colour nor a type, that leaves 3 077 of 455 608 stars (274 of 4 735 hosts) with
no radius; they are drawn at the Sun's, and their card gives none.
The card says which it is: "Radius 0.141 solar radii" for a published one, "~0.10 solar radii,
from colour and brightness" for a derived one, two figures because colour does not give three.
A giant drawn at its size can be wider than its system, so systemFramingDistanceAu also makes
room for the star, and the controls' closest approach is now three of the star's radii where
that is more than the old 0.05 AU. 23 211 stars are drawn wider than 3.6 solar radii, which put
0.05 AU inside three of their radii, and a zoom would have carried the camera through the
surface of the largest. The Sun keeps 0.05 AU. starMarkerRadiusAu and the renderer's innermost
axis, which only it read, are gone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A star's luminosity was its absolute magnitude plus a bolometric correction read off its spectral
type, with its magnitude taken as V whatever band it was in. Gaia classifies none of its stars, so
all of its 379 000 got the Sun's correction, and their G was read as V: TRAPPIST-1 came out at a
seventh of its luminosity.
The dwarf sequence spectral.ts already reads types off (Pecaut & Mamajek 2013, table 5, online
version 2022.04.16) now carries its effective temperature, bolometric correction to V and Gaia
G-V columns, and dwarfSequenceAtColor interpolates them at a colour, in the colour's own system.
luminositySolar uses it wherever the star's colour is inside the table: the G magnitude is
carried to V, then corrected. Only without such a colour does it fall back to the spectral type,
as before.
Against the archive's own st_lum for the 1 449 hosts that are catalogue stars, the median error
goes from 0.038 to 0.022 dex and the 90th percentile from 0.292 to 0.115 dex; within a factor of
1.5, 84.5 % -> 93.8 %. For the 572 hosts Gaia describes: 90th percentile 0.332 -> 0.073 dex,
81.8 % -> 97.0 % within a factor of 1.5. Barnard's Star with no type now reads 0.0029 L_sun
against 0.0035 published, and TRAPPIST-1 7.3e-4 against 5.5e-4 (Agol et al. 2021).
Across the catalogue, 311 255 of 455 608 stars move by more than 10 % (median ratio 0.90: a
G-type star's G is 0.16 brighter than its V), and so do the hosts of 3 158 planets, whose
equilibrium temperatures follow as the fourth root.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The README listed the four columns stars-meta.bin had before a81dd49 added the photometry byte
(magnitude band, colour system, whether the distance is Gaia's) and the distance's relative
error.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The derived equilibrium temperature balanced starlight alone. For the widest orbits that is less
than the 2.7 K every body in space is held at by the cosmic microwave background. The planet 7 493
AU from 2MASS J21252752-8138278 read "Equilibrium temp. 1 K", and the one 19 000 AU out from UCAC4
328-061594 would have read 0 K.
equilibriumTemperatureK now adds the background as a second source in the same balance, T^4 =
T_star^4 + (2.7255 K)^4 (Fixsen 2009). Inside a few hundred AU of any star it changes nothing that
shows: Earth, Mars, Jupiter and Neptune reproduce their published values as before. Across the
6 354 exoplanets, the rounded temperature changes for 11, all on orbits of 350 AU or more. Ten go
from 0, 1 or 2 K to 3 K, from VHS J125601.92-125723.9 b at 350 AU to UCAC4 328-061594 b at 19 000
AU, and 2MASS J22501512+2325342 b at 518 AU goes from 4 to 5 K. On the dev server, the detail
pages of the 7 493 AU planet and of GJ 900 b read "Equilibrium temp. 3 K".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The detail page and the system view's object card truncated the name and the line under it with
an ellipsis. For a designation, the part that went is the part that identifies it: the audit
measured "2MASS J21252752-8138278 b" at 288 px in a 256 px heading, shown as
"2MASS J21252752-81382…", and the eyebrow at 273 px, shown as "EXOPLANET · 2MASS J21252752-8138…".
Both lines now wrap (wrap-break-word) in both places. On the dev server the heading for that
planet is two lines, 45 px high, with its scroll width equal to its 256 px box, and the eyebrow
fits in 256 px.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2 127 stars, Sirius among them, read "Spectral type A0m...", in search rows, route options and
the star card. The mark is Hipparcos's own: every one of the 3 803 HYG rows ending in it has a HIP
number, and the catalogue ends a classification it does not print in full with it. The app
truncated nothing, but the text reads as if it had.
fetchStars now trims a trailing "..." from HYG's spectral types. The dictionary in
stars-index.json goes from 3 056 distinct types to 2 883, and from 598 dotted ones to none. The
names, sources and source indices are unchanged, and so is every other column of stars-meta.bin
bar the type indices. gzip -9 of the index goes from 4 015 865 to 4 015 205 bytes.
build.ts validateStars now fails a catalogue with any type ending in "...". Run with the trim
removed, the ETL failed on 2 127 such types. Two ETL runs from cache wrote identical
stars-meta.bin and stars-index.json.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
383 695 of the 455 608 stars read "Spectral type Unknown", every Gaia star among them, although
380 884 of them carry a colour. The card now gives those the type of the dwarf whose colour is
nearest, in the colour's own system, marked as an estimate: TRAPPIST-1, BP-RP 4.90, reads
"Spectral type ~M8, from colour", which is how it was classified (M8 V). A star with neither a
type nor a colour gets no subtitle rather than the literal "Unknown".
The colours are Pecaut & Mamajek's mean dwarf sequence (2013, ApJS 208, 9, table 5), as Mamajek
maintains it online (version 2022.04.16, which carries Gaia BP-RP), from B0 to M8.5. B-V cannot
tell O types apart (the whole sequence spans 0.03 of it), and BP-RP turns back past M8.5 and is
tabulated only from B9. A colour outside the table gets no estimate: 186 BP-RP and 41 B-V colours,
among them the blue Gaia DR3 5612323414549657984 at BP-RP -0.15.
380 657 stars get an estimate: F 97 840, G 144 327, K 105 984, M 29 142, A 3 272, B 92. Checked
against catalogued types:
- 19 433 HYG dwarfs with a B-V: 74.3 % within two subclasses, 95.5 % within five.
- 146 Gaia exoplanet hosts the archive types as dwarfs, from BP-RP: 88.4 % within two
subclasses, 98.6 % within five.
The estimate assumes a dwarf, so a giant reads later than it is (Pollux, K0 III at B-V 0.99,
reads K3). It also ignores reddening. The comment says both.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The readout named Hipparcos, Yale and Gliese while 378 775 of the 455 608 stars are described by
Gaia DR3, printed one "Magnitude" for G and V alike, and gave every distance to the parsec. The
band cannot be read off the star's source, which records whose position it has: 62 002 stars Gaia
places keep HYG's V and B-V, and the 575 hosts renamed after their planets are Gaia's, in G.
stars-meta.bin gains two byte columns, 14 to 16 bytes a star (6 378 512 to 7 289 728 bytes; gzip
-9 2 804 049 to 3 110 991). One holds the magnitude's band (V 76 555 stars, G 378 744, none 309,
whose magnitude is a stand-in), which colour the colour index is (B-V 75 203, BP-RP 376 703), and
whether the distance is Gaia's parallax. The other holds the distance's relative error as its
square root in 255ths: a step is 0.08 % of distance at 1 %, 0.35 % at 20 %, and 100 % is the top.
encode, decode, BYTES_PER_STAR_META and the build.ts round trip cover both; no workflow reads the
format.
Where the errors come from:
- Gaia rows keep the parallax_error their query already fetched: median 0.3 %, 90th percentile
1.2 %, at most 20 %, the query's own cut.
- A HYG star at Gaia's distance takes the cross-match's parallax_over_error, and a star merged
into a Gaia entry keeps that entry's error with its position.
- The 3 067 Hipparcos stars that keep their Hipparcos distance, Rigel, Deneb and Alnilam among
them, take e_plx from van Leeuwen's 2007 reduction: a new cached query of
public.hipparcos_newreduction on the ESA archive, whose 117 955 rows HYG's distances invert.
- The archive's stars take sy_disterr1/2 from pscomppars, in a query and cache file of their own
so the composite rows already cached were not refetched.
- 439 distances have no published error: 357 Gliese rows and 82 archive hosts.
Of the errors, 392 786 are 1 % or less and are not printed; 61 156 print as "117 ± 12 pc" to the
distance's own digits; 1 196 between 20 and 100 %, and 30 past it, print as the range the
parallax gives, since a symmetric error in parallax is a lopsided one in distance.
The star card (measured on the dev server) now reads, for example:
- Rigel: 265 ± 23 pc, V 0.18, B-V -0.03, source HYG.
- Deneb: 433 ± 60 pc.
- Alnilam: "476 pc to 833 pc" (Hipparcos 1.65 ± 0.45 mas).
- Gaia DR3 5612323414549657984: 111 ± 2 pc, G 4.63, BP-RP -0.15, source Gaia DR3.
- Proxima Centauri: 1.30 pc, V 11.01, source "HYG, Gaia DR3 distance".
- TRAPPIST-1: G 15.62, BP-RP 4.90, source Gaia DR3.
- Kepler-186: V 15.14, source NASA Exoplanet Archive.
The neighbourhood's subtitle reads "Gaia DR3 378,775 · HYG 73,556 · NASA Exoplanet Archive
3,277", counted by the catalogue describing each star.
build.ts validateStars now fails a catalogue with more than 1 000 stars without a band (309
today) or without a distance error (439). Dropping G from the Gaia rows gave 379 040 without a
band, and dropping their parallax_error gave 441 216 without an error; both runs failed.
Decoding the catalogue in Node took a median 29 ms before and 24 ms after (nine runs each, within
noise). In the app, five cold boots gave a 654-786 ms long task after the data landed and the
HUD at 1.83-2.07 s.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
After matching, 4 237 planets still had no star: their hosts are too faint for either Gaia
query (fainter than G 12 past 50 pc) or too far (past 250 pc), Kepler-186 at 177.6 pc among
them. fetchExoplanets now adds one star per such host from the archive's own figures, 3 277
of them, whenever the archive gives a position and a distance:
- position carried back from J2016, where the archive publishes it (741 of the 746 matched
hosts moving over 100 mas/yr sit nearer their star carried back, a median 0.11" against 3.47"
as published), and placed at sy_dist;
- magnitude in V (2 999 hosts), else Gaia G (13), the band the catalogue's Gaia stars are
already in; 265 have neither, 128 KMT, 95 OGLE and 32 MOA microlensing hosts at a median
6.2 kpc among them, and take the ETL's faint stand-in of 15;
- colour as B-V from the archive's B and V, else from st_teff through a new
temperatureToColorIndex (Ballesteros 2012, inverted; the Sun's 5 772 K gives 0.65), else
left to st_spectype;
- ids from 1 070 000 000, past Gaia's two ranges and under the 2^30 validateStars enforces;
source "exoplanet-archive".
Hosts past 250 pc are included: 399 of the 3 277 are within 250 pc, 1 962 between 250 pc and
1 kpc, 916 beyond. The drawn budget still chooses what is drawn (70 000 of 455 608).
Planets with a star: 2 090 -> 6 327 of 6 354. The other 27 have no distance in either table
(Luhman 16 A, mu2 Sco, PSR B1620-26 among them). Systems in the Solar Neighbourhood readout:
1 450 -> 4 736. validateExoplanets now refuses a catalogue where fewer than 99.5 % of planets
have a host (measured 99.58 %); dropping the added stars fails it at 2 090, and dropping the
composite fill at 6 227. No added star sits within an arcsecond of a catalogue star
(validateMerge's twins stay at 23); 5 of the 399 within 250 pc have one within a minute of arc,
VHS J125601.92-125723.9 (archive 12.7 pc) 5.8" from a Gaia entry at 21.2 pc the likeliest
duplicate.
In the app, searching TRAPPIST-1 or Kepler-186 and picking the star enters a system with its
seven and five planets drawn. gzip -9: stars.bin 5 030 741 -> 5 067 864 B, stars-meta.bin 2 784 614 -> 2 804 064,
stars-index.json 4 003 584 -> 4 015 882, exoplanets.json 342 467 -> 446 532 (host parameters,
both commits).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The archive's default-parameter rows leave sy_dist blank for 100 planets, TRAPPIST-1's seven
among them, so the matcher could not place a host the catalogue had drawn since the Gaia
nearby query (Gaia DR3 2635476908753563008, 12.47 pc). fetchExoplanets now also reads the
Planetary Systems Composite table (pscomppars) and fills a default row's blank host cells from
it. Only host columns: a composite row takes each column from its own reference, so orbits
still come from the default row alone, one fit per planet. Where both tables give a distance or
a position they agree on all 6 225 and 6 352 rows.
Three hosts the catalogue holds by name failed the distance ratio test because the archive's
sy_dist, from TICv8, contradicts its own parallax: Lalande 21185 (GJ 411) 5.68 pc against
392 mas, Luyten's Star (GJ 273) 5.92 against 263 mas, Struve 2398 B (Gl 725 B) 6.84 against
285 mas. resolveHostStarId now takes the archive's parallax (sy_plx) as a second distance the
ratio test accepts. It is a second chance, not a replacement: 47 of 5 959 systems disagree past
the tolerance, and for faint far hosts the inverse parallax is the worse figure (K2-238: 538 pc
by sy_dist, 6 779 by parallax).
Planets on a catalogue star: 2 071 -> 2 090 of 6 354. The 19 gained are TRAPPIST-1 (7),
GJ 273 (2), GJ 411 (2), Gl 725 B, HD 62509 (Pollux), K2-65, TOI-2267 A and B, and two
brown-dwarf hosts, 2MASS J02192210-3925225 and DENIS-P J082303.1-491201. No planet lost or
changed its host.
A matched host whose catalogue name is a bare Gaia designation now takes the archive's host
name, so search finds TRAPPIST-1, Teegarden's Star, TOI-700, LP 791-18 and K2-18: 575 stars
renamed. validateExoplanets refuses a host that is still only a designation, and the star
assets are rewritten after the exoplanets for that reason; stars.bin and stars-meta.bin are
unchanged. TOI-2267 A and B both land on one Gaia entry, which takes the name TOI-2267 A.
Each planet also carries its host's radius, effective temperature and luminosity (10^st_lum),
and a mass for 6 344 planets instead of 5 474, from the default row where it gives one and the
composite table otherwise, for the star-physics step.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
OpenNGC keeps the objects no NGC or IC number covers in a second file, addendum.csv, with the
same 32 semicolon-separated columns as NGC.csv. The ETL only ever fetched NGC.csv, so the
brightest deep-sky object in the sky, the Large Magellanic Cloud (V 0.29), was missing while
the Small one was drawn, and so were the Pleiades (M45), the Hyades, the Horsehead and the
Coalsack. fetchDeepSky now fetches the addendum beside NGC.csv, caches it as
openngc-addendum.csv, and runs its rows through the same loop.
Of its 64 rows, 27 pass the existing filters: all 22 named or Messier rows except M40, which
OpenNGC types as a double star, and M102, typed as a duplicate of M101; plus seven anonymous
open clusters brighter than V 9 (H05, H20, H21, Mel071, Mel101, Mel105, MWSC3171). deepsky.json
goes from 463 to 490 objects (galaxies 73 -> 85, clusters 294 -> 306, nebulae 96 -> 99), with
no existing record changed. Messier coverage goes from 106 to 107 of 110. 340 of the 490 have
a distance: the Pleiades 135.8 pc and the Coma Star Cluster 85.9 pc from their parallaxes;
the Hyades and the Local Group dwarfs honestly have none.
validateDeepSky now requires the Andromeda Galaxy and the Small Magellanic Cloud from NGC.csv,
the Large Magellanic Cloud and the Pleiades from the addendum, and at least 107 Messier
objects. Both checks were run against the real catalogue with the addendum removed: the ETL
fails on "Deep-sky object ESO056-115 is missing", and with the required list emptied, on
"Only 106 Messier objects were produced".
In the app, on the dev server: 490 sprites. The Pleiades sprite lies 0.00182 deg from Alcyone
(0.00183 deg from the published coordinates); the LMC 18.4209 deg from Canopus and 26.8215 deg
from Achernar (18.4209 and 26.8215 published); the Hyades 2.2591 deg from Aldebaran (2.2591).
The twelve labelled deep-sky objects now open with the Large Magellanic Cloud and the
Pleiades and include Brocchi's Cluster, in place of h Persei, chi Persei and NGC 2516.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Gaia's G < 12 cut took a quarter of what lies within 10 pc: the red,
brown and white dwarfs most of the neighbourhood is made of, which the
map only had where Gliese happened to list them. Teegarden's Star was
missing, and with it its three planets' host.
A second query fetches the complement out to 50 pc: G >= 12 or no G,
parallax > 20 mas, with pmra/pmdec so the J2016 -> J2000 propagation
applies. The quality filter was chosen by counting. Parallax over error
> 5 keeps 39 751 of the 39 764 sources that pass the floor, but the Gaia
Catalogue of Nearby Stars (GCNS; Gaia Collaboration, Smart et al. 2021)
rejects 11 752 of them as spurious: median G 20.2, astrometric excess
noise 5.4 mas against 0.15 for the ones it keeps, 6 933 toward the
Galactic centre. So the query joins the GCNS main table (EDR3 astrometry
and source ids, which DR3 carries unchanged) and keeps 28 012 rows; the
error cut stays and costs no GCNS source, brown dwarfs included. The
query has its own row-count floor (28 012) and the row-limit cap, and is
ordered by (phot_g_mean_mag, source_id); a second ETL run reproduced
stars.bin, stars-meta.bin and stars-index.json byte for byte. Its ids
start at 1 050 000 000, clear of the main query's and under 2^30, which
V8 keeps unboxed: numbered from 2 000 000 000 they made the app's boot
task 230 ms longer (medians of five interleaved runs, 1.41 s against
1.18). validateStars now refuses an id outside 0 to 2^30.
The Gliese entries these stars duplicate were not folded: HYG carries
them with positions off by up to a minute of arc and photometric
distances, so they missed the 15" tolerance or failed the distance test.
Their proper motions, which Gliese measured well, give them away:
isSameStar now takes two entries moving within 20 % of each other as one
star up to 60" apart, whatever their distances, brightness still
permitting. Of the 602 Gliese-only rows left without a counterpart, 253
have such a Gaia entry; with every entry shifted a quarter degree, none
does. fetchStars passes HYG's motions only for rows without Hipparcos
astrometry: given them too, 15 Hipparcos stars took a co-moving
companion's Gaia entry and the cross-catalogue pairs under an arcsecond
went from 23 to 35. Of HYG's 1 200 stars fainter than V 12 within 25 pc,
1 024 now sit on a Gaia position (325 before); of the 176 left alone, 43
still have a Gaia entry 3-60" away (218 without the motion rule), some of
them real companions.
Measured on the rebuilt catalogue, against the GCNS (sources with
parallax > 100, 40 and 20 mas):
within 10 pc 336 -> 372 (GCNS 312; the map adds 60 HYG-only stars)
within 25 pc 3 652 -> 5 560 (GCNS 5 111)
within 50 pc 13 702 -> 40 916 (GCNS 40 231)
452 331 stars (+27 214). Proxima, Barnard's Star, Wolf 359, Rigel, Deneb
and Alnilam are all present by name; Teegarden's Star is Gaia DR3
35227046884571776 at 3.83 pc and hosts its three planets. Luhman 16 is
not in Gaia DR3 with a parallax (5353626573555863424 has a two-parameter
solution) and stays absent. 94 more exoplanets find a host (2 071), none
changes host. TRAPPIST-1 is now drawn (Gaia DR3 2635476908753563008,
12.47 pc) but its planets are not yet matched to it.
Merge gate, ceilings unchanged: HYG rows without a Gaia counterpart
12 352 -> 11 554 (ceiling 15 000; 10 886 before the naked-eye stars),
cross-catalogue pairs under an arcsecond 23 -> 23 (ceiling 100). The
gate's comment now accounts for the survivors by magnitude.
gzip -9 sizes against the catalogue before both changes: stars.bin
4 711 922 -> 5 030 741 B, stars-meta.bin 2 576 213 -> 2 784 614 B,
stars-index.json 3 724 855 -> 4 003 584 B (+806 KB, 7.3 %). Boot on the
dev server, five interleaved cold runs: the task that indexes the
catalogue after the data lands, median 1 072 -> 1 182 ms; HUD shown,
median 2 622 -> 2 716 ms. This machine measured 0.82-1.30 s for the
same baseline task today, above audit #25's 627-843 ms. The drawn-star
budget is unchanged.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The 250 pc cutoff took 1 543 of HYG's 8 920 stars of V 6.5 or brighter:
1 339 that both surveys put past it and 204 HYG has no distance for. Five
of the fifty brightest stars in the sky were gone (Rigel, Deneb, Alnilam,
gamma-2 Vel, Wezen), and Naos, Sadr, Aludra and Arneb with them, while
11th-magnitude Gaia stars at the same distance were drawn. The cutoff
bounds a download, not what the sky shows.
placementDistancePc now keeps a star of V 6.5 or brighter at any distance,
at the better of its two distances as before: Gaia's where the archive's
Hipparcos cross-match has a usable parallax, Hipparcos's otherwise. Gaia
saturates on the brightest, so Rigel (264.6 pc), Deneb (432.9), Alnilam
(606.1), Wezen (492.6), Naos, Sadr, Aludra and Arneb sit at their
Hipparcos distance. 1 502 HYG rows come back, 165 of the 206 without a
Hipparcos distance among them because Gaia measured them; 36 fold into a
Gaia entry. 41 naked-eye stars stay out because neither survey gives them
a distance: beta Phe, Polis, Mu Cep, Rho Cas, Eta Car, Alp Cam, Phi Cas,
Chi Aur, Psi-1 Aur, theta-1 Ori, Omi-1 Cen, 66 Ori, 16 Sgr, 10 Sge and 27
HD stars.
Measured on the rebuilt catalogue: 425 117 stars (+1 466), 8 301 of them
past 250 pc (+1 466), the same 336 within 10 pc and 3 652 within 25 pc.
Merge gate: 12 352 HYG rows without a Gaia counterpart (was 10 886; the
ceiling of 15 000 is unchanged, and its comment now counts the naked-eye
stars among the survivors) and the same 23 cross-catalogue pairs under an
arcsecond. Five more exoplanets find their host: HD 81817 b and c,
HD 158996 b, HD 208527 b, HD 220074 b. gzip -9 sizes: stars.bin
4 711 922 -> 4 728 315 B, stars-meta.bin 2 576 213 -> 2 587 977 B,
stars-index.json 3 724 855 -> 3 731 763 B.
The brief behind this also asked to cut once on the best distance, which
would drop the 6 835 stars Hipparcos puts inside 250 pc and Gaia outside.
Those already sit at Gaia's distance (4eb61ff), so nothing is misplaced,
and they stay.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The off-screen rule for clicks was described in 3f0abf8 and in the pull request, but only its
comment was committed: pickAt still let the slop reach past the frame. The mutant that was said
to catch it matched nothing, and an unrelated flaky test failed instead. The frame test is now in
pickAt, before the slop, and its test fails without it (star at NDC 1.01, click at 0.995).
Venus, Uranus and Pluto turned forwards: Horizons states a retrograde spin twice, by a negative
rate and by an obliquity over 90 degrees, and both were applied. The period's sign is now used
only when no obliquity is known. Measured on the live markers, spin axis against orbit normal is
cos(obliquity) for each: Venus -0.999, Uranus -0.135, Pluto -0.494, Earth 0.917.
Moons listed as rates rather than "Synchronous" drifted about 5 degrees an orbit and Titan did not
turn: every moon is now locked at its Kepler period. Pluto's obliquity comes from IAU WGCCRE 2015,
Horizons gives none.
Also:
- the star's light is white at pi, not a warm 2.2 that left the photographs dim;
- procedural textures are 128x64, not 512x256 that froze the main thread ~60 ms a body;
- Io, Pluto, Titan and Deimos lose their "maps", which were disc photographs with black sky;
- an exoplanet with only a mass gets a radius from it (M^0.55, capped at Jupiter), not Earth's;
- the clock knows when it has left the present even once back at real time, so the date and
"Back to now" stay up.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every body in the system view was an unlit sphere wearing a 32 by 16 pixel
procedural texture — the size chosen when a marker was a few pixels across and
what survived was its average colour. The thirteen real photographs in
`src/assets/textures/bodies/` were used only by the detail page. So Mars was a
pale grey ball with invented polar caps while its own NASA mosaic sat unread in
the repository, and nothing had a day side or a night side.
Each marker now takes its own photograph where one exists, at the size the
detail page uses, and the derived texture only where none does — the five moons
no probe mapped, and every exoplanet, none of which has ever been imaged. The
material is lit, and the light is a point at the star, so each world shows the
terminator where it really falls.
The light does not fall off with distance. Under the inverse square that real
light obeys, Neptune receives a thousandth of what Mercury does and reads as
black; the map is a set of worlds to look at rather than a light meter, so each
is lit as a photograph of it would be. That is the same concession the pixel
floor makes for size, and it is only about brightness: the *direction* is real.
Spheres are 32 by 24 rather than 16 by 12, since at true scale a body is drawn
anywhere from a pixel to the whole frame and the old silhouette was visibly
faceted at the near end.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The orbits and the rotations are both functions of a date, and the only date the
map ever asked for was this instant. So a view built on propagated ephemerides
showed a still picture: Earth turns 15 degrees an hour and takes a year to go
round, and a reader watching for a minute saw nothing move at all.
`TimeStore` is that date, at a rate the reader sets: real time, an hour a
second, a day a second, a month a second. It is read once a frame rather than
held in a signal — it changes continuously, and a signal changing sixty times a
second would ask the whole HUD to re-render for a number nothing is watching.
Changing the rate re-anchors rather than rewinding, so speeding up and slowing
down never jumps the sky, and "Back to now" returns to the world's own time.
Measured in the app, three seconds of watching in the Sun's system:
| rate | sky elapsed | Earth turned | Jupiter moved |
|---|---|---|---|
| real time | 0 | 0 | 0 |
| 1 h/s | 3.0 h | 45.12 deg | 0.0009 AU |
| 1 d/s | 3.0 d | (three full turns) | 0.0222 AU |
45.12 degrees in three hours is 15.04 an hour, which is Earth's own sidereal
rate, and Jupiter's 0.0222 AU in three days is its own orbital speed.
The rates are radio buttons, not toggles: they are one of four, and the native
control carries that to a screen reader and to the arrow keys with no script.
The date joins the strip only while the clock is running faster than the world,
since at real time it is today's, which the reader's machine already says.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The view had one rotation in it — the planet on the detail page, at 0.08 rad/s,
a number with no source. Nothing in the system view turned at all.
The data was already on disk: every cached Horizons page carries how its body
spins, in one of five forms. The rate in radians per second is preferred where
it appears, because it is signed — that is how Venus and Uranus are known to
turn backwards — then a period in hours or days, then the `9h 55m 29.711 s`
the giant planets use, and finally the word every major moon here carries
instead of a number: Synchronous. A tidally locked moon's day is its orbit, so
Kepler supplies it from the elements already parsed and the parent it goes
round.
Seventeen of the eighteen bodies come out within 1% of their published period —
Earth 23.934 h, Jupiter 9.925 h, Venus -5832.5 h, Io 42.5 h, Callisto 400.5 h.
Titan is the exception: its page states no period at all, so it is left still
rather than turned at an invented rate.
The axis is the orbit normal tilted by the obliquity about the orbit's
ascending node, which is where an obliquity is measured from and the only line
in the orbit the elements name. The phase at the epoch is published for none of
these bodies, so the face turned toward the camera is not a claim; the rate and
the direction are.
At true rates nothing is visible moving — Earth turns 15 degrees an hour. A
clock the reader can run faster is the next piece, and the audit asks for it
anyway.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Three changes to what the system view claims, all of them the same claim: that
the sizes on screen mean something.
**The halo is gone.** It was a sprite sized against the arrival frame — 1.12 AU
for the Sun — so it stayed that wide as the camera closed in and ended up a flat
gradient filling the screen, over the photograph it was meant to dress. It
existed to keep the star visible at a framing that holds the whole system, which
is now handled in pixels instead.
**Bodies are drawn at their own radius.** The old marker size was exaggerated
and scaled to the system span, and clamped: Jupiter and Ganymede both ran past
the ceiling and were drawn at one radius, so every moon orbited inside its
planet, and Phobos and Triton sat entirely within Mars and Neptune. True scale
needs no rule against that — physics already puts a moon outside the planet it
orbits. What it costs is visibility at the arrival framing, where every body is
sub-pixel, so the scene floors each marker at 3 px on screen and holds a moon to
half its planet's drawn size. Measured in the Sun's system: at arrival, planets
3 px and moons 1.5 px, against 3 px for everything before; at Jupiter, the
planet 10.8 px at scale 1 with the Galilean moons on their orbits outside it.
The Sun is drawn at its own radius too. Every other star keeps a size derived
from its innermost orbit, because no stellar radius reaches the app — Gaia's
`radius_gspphot` is the obvious next fetch.
**A click cannot enter a system that is not on screen.** The picker tested depth
but not the frame, and a star's hit area is its drawn size plus a slop, so a
click in the last pixels of the view could fly into a system outside it, with
nothing on screen to explain where it had gone.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Both sides added a test beside the other in the dock's spec: the panel's own
departure guard here, the give-up wording on main. Both kept, and the offer test
carries the `least` the route answer now has.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The span went to `distanceRings` as the target's straight-line distance from the
Sun, but a ring of radius r passes within |r - p| of the view's centre, where p
is how far out that centre is *along* the galactic plane. For a target above the
plane the two differ by its height, so the band was centred on a radius no ring
has — and `ringLabels` picks its bearing by comparing its own in-plane distance
against the innermost ring, a comparison the new first ring quietly broke.
Two comments and a constant, from the same review. A frame short of the survey
edge gets its callout only when its last ring overshoots it: 245 pc does, 235 pc
does not, which is now a test rather than a sentence. The ring count can reach
16, not 14, now that the span need not start at the Sun. And a ring label was
measured as 135 px of star name when "50 pc" is a third of that, which rejected
rungs a hand's breadth clear of the name: RING_LABEL_REACH_NDC, 0.23, is the
widest of them — "1.5 kpc" with "Survey edge" under it.
Three mutants, three caught. Measured again in the app: unchanged for a star in
the plane (4 labels at 20 pc above it, 7 at 2 pc), and the rings now follow the
plane for one 195 pc above it rather than ringing a place the grid does not
reach.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The fixture built for "exactly MAX_VISITED stars reachable" was 338 short: its
random cloud leaves clumps the departure never reaches (the LCG gives 12 212
distinct positions for 39 999 stars), so the search settled 39 662 and the
pre-fix code answered `gaveUp: false` too. The test could not fail on the code
it was written to pin — and the mutant that seemed to prove otherwise was
failing to compile, not failing the test. It is now a line of 40 000 a parsec
apart with the island off the line: settled 40 000 exactly, 115 ms, and the
pre-fix code does report a give-up. Both mutants now compile and are caught.
The give-up cap also has to hold while the bisection has earned nothing: the
exception added for that case had no bound at all, so a search could spend the
resolution's own eight full-budget probes — about 17 s of "Plotting…" — where
two used to cost 4 s. Bounded at five. On the repo's crowded-knot fixture:
1.9 s for the unearned ceiling figure with the old cap, 7.2 s for a range the
bisection earned, and five probes is where that lands.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Gating it on `jobsQuery` switched it off for every override that *widens* the
query — which is the only way to fill `select top N` at all. `ETL_GAIA_MAGNITUDE_LIMIT=14`
asks for 500 000 rows, the sky holds more, and the answer is the limit rather
than the filters: exactly what the tripwire is for, and it no longer fired. It
now reads the row limit itself, so only a deliberately smaller slice is silent.
Measured with a synthetic answer of exactly 500 000 rows in the cache, under the
key the widened query hashes to: refused. With the `jobsQuery` gate back, the
same run keeps 500 000 Gaia stars and goes on to publish them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
`canPlot()` guarded the Plot button and not `raiseTo`, which is the other way
into `plot()`. So with a departure typed but never chosen, clicking "1.8 pc
would reach." moved the range control and plotted nothing: the panel then read
"No route at this range. 1.8 pc would reach." beside a control already set to
1.8. The offer carries the same `disabled` as the button, since it is the same
request by another route.
And the departure guard is trimmed, as the scene trims the same text before
offering matches for it: one space in the field left it looking empty, with no
suggestions to pick from, and Plot dead for no reason on screen.
Measured in the app, from inside Barnard's Star with Sirius as the destination:
offer enabled with the field empty, disabled once "Sol" is typed and never
chosen — a forced click then moves nothing — and enabled again when the field is
cleared, where it raises the range to 2.40 pc and plots 7 jumps.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The rings are centred on the Sun and the frame need not be. Sizing their step
from how far the frame reaches — 210 pc for a star at 190 with the camera 20 pc
back — gives 20 pc rings at 180 and 200, both outside a frame 19 pc deep, so a
view away from the Sun still had no ring on it and no ladder of labels either.
`distanceRings` now takes the span the frame covers rather than its far edge,
and the step is a fifth of that: 5 pc rings from 165 to 210 for the same view.
Measured in the app, centred on a star 187 pc out in the galactic plane: 4 ring
labels drawn 20 pc above the plane and 7 from 2 pc, against 1 and none before.
The clearance was a radius around the anchor, and a label is a line of text
hanging 135 px to one side of its anchor: at 0.065 NDC apart, past the radius,
"50 pc" printed inside "Alpha Centauri". It is now tested against the span the
name occupies, on the side it hangs, with the radius kept for the pair whose
text runs the other way.
Also from the review: the ladder in the clearance test was built at exactly the
constant it tests, so 1057 of 2000 camera poses would have decided it by float
round-trip error — the rungs now sit 0.02 either side of the rule. And two
comments that were wrong: a frame one step short of the survey edge does get its
callout, and CSS2DRenderer hides a label behind the camera rather than drawing
it at the page edge.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The panel printed "No route at this range." beside the range it was offering —
which is the sentence this branch exists to stop it printing. It was gated on
there being no offer, and a search that gives up usually has one: Sol to
HD 120147 at 4.5 pc spends the budget, offers 4.83 pc, and says there is no
route where a 71-jump route exists. The wording now follows the search at the
range that was asked for, and nothing else.
That needs the two give-ups kept apart, so `least` travels beside `gaveUp` to
the panel: one says the asked range was not searched out, the other that the
search for a range that would work was. HIP 69445 at 3 pc — asked-range search
exhaustive in 44 ms, ceiling probe out of budget — used to read "Too many stars
to search at this range." and now reads "No route at this range.", with nothing
claimed after it.
Two more from the same review. The budget flag was read off the settled count,
so a search that proved a dead end with the last star it was allowed reported a
give-up; it now records why the loop stopped. And the bisection's cap could fire
before a single probe had narrowed anything, leaving the ceiling route's own
longest hop as the answer: star 1000115173 at 3 pc was told to go to 8.00 pc,
the control's maximum, for a crossing that works at 6. It now offers 6.93.
Measured in the app, all three: "Too many stars to search at this range.
4.90 pc would reach.", "No route at this range." alone, and 7.00 pc in place of
8.00. The duplicated dead-end test now asks the question it was named for —
exactly the budget's worth of stars reaching each other and none of them the
destination — and each fix kills its own mutant.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Three things this got wrong. The direction: truncating Gaia leaves the HYG rows
whose counterpart it dropped without one, so survivors rise — 10 886 today,
12 711 at half the rows, 16 258 at a third — which the comment claimed was the
other way, and which decides whether the 15 000 ceiling can be leaned on at all
(it catches a truncation past about two thirds, and nothing shallower).
The throw: `fetchStars` catches everything a source throws and skips it, so a
truncated CSV was reported as "the archive was unreachable" one step after
`writeStarAssets` had already overwritten the published catalogue. Marked with
`GaiaAnswerError` and rethrown there, so an answer that cannot be worked with
fails the run where it happened. Measured end to end in a throwaway working
directory, 300 000 rows in the cache: fails, names the cache file to delete,
assets untouched. With the rethrow taken back out again: assets written, then
"the archive was unreachable".
The row limit: `rows.length >= ROW_LIMIT` is true for every reduced
ETL_GAIA_ROW_LIMIT, so the tripwire fired on exactly the deliberate slice the
override exists for — and told the operator to raise it. Gated on the same flag
as its neighbour. `ETL_GAIA_ROW_LIMIT=20000` now runs through; without the gate
it dies on the limit it was given.
Also: the row floor names the one cache file it is about rather than a glob that
takes the Hipparcos cross-match with it, and says an edited query is a third
reason it can fire — DEFAULT_QUERY_ROWS now sits under the query it counts.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Both make a search spend its whole 40 000-star budget, twice over in the bisection, and the
CI runner timed out at the default five seconds. The crowds are now indexed in cells sized for the
ranges asked of them, as the real catalogue is, and the two tests carry their own 30 s timeout.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
From the review of #24. Two defects, both in a real browser.
The panel keeps its entries across a trip to another tab, but still replayed the acquire wipe on
the way back, and for the 380 ms that runs, its clip path swallows clicks: type "Siri", leave for
Readout, come back and click the Sirius suggestion, and the click lands on the star field behind
it — measured, the element under the pointer is the canvas, and the field stays "Siri". The wipe is
gone from this one panel: it is not acquiring anything it did not already have.
The departure field fell back to the star the view is in whenever nothing had been chosen, text in
the field or not. So a field reading "Sol" that was never resolved plotted from Barnard's Star:
"1 Barnard's Star, 2 Sol, 3 Sirius", the panel naming one departure and the route leaving from
another. Text nobody chose is no longer a departure, and the button waits until it is one or the
field is empty again.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
From the review of #19. The rings are distances from the Sun, but their step was taken from
`effectiveDistance`, which under the plan view means the extent of the frame rather than how far
the camera is from the Sun. Centred on a star 200 pc out and flipped to 2D, the grid became rings
of 2 to 20 pc: not one of them on screen. The step now comes from where the view is centred plus
how far the camera is orbiting it, which is the same distance under either projection.
The set was also rebuilt while the grid was hidden, and every rebuild disposes the rings and
builds every vertex again; it now happens only while the grid is drawn.
The ring labels went straight to the overlay: never culled to the frame, and free to land on a
star's name. They now have to be on screen and clear of the names already placed, by half the
separation two names keep — they are a ladder up one ray a twentieth of the screen apart, and
holding them apart from each other would take "Survey edge" off the map.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The route search stops after MAX_VISITED stars and returned null, which everything downstream read
as "the catalogue holds no chain". On the real catalogue that was wrong for real questions: Sol to
HD 120147 (136 pc) at 5 pc is 50 jumps, and the panel said there was no route. The budget also sat
under what the shipped catalogue needs, so it is now 40 000 rather than 20 000: both that route and
a star at 170 pc are found, and Sol to HD 2626 at 6 pc, which used to be refused after 4.7 s, plots
56 jumps in about 2 s.
A search now reports whether it gave up. The range search no longer counts a give-up as proof that
nothing routes below it — that is what reported ranges up to 29% too wide — and it stops after two
of them, since those are the probes that cost the most and settle the least: for HD 2626 at 3 pc it
offers 5.92 pc in about 4 s, against 6.13 pc in 4.7 s. At the panel's widest range the refused route
and the range search are the same question, so it is asked once. Where nothing can be said, the
panel says "Too many stars to search at this range." rather than claiming there is no route.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The merge gate asks whether Gaia contributed any stars, never how many. The TAP service truncates
on its own timeout and still serves a well-formed CSV with a 200, ordered by magnitude — so a half
answer is the bright half, which is the half HYG overlaps. Every gate passes: Gaia stars are
present, HYG survivors go down rather than up, unmerged twins can only fall. The weekly job would
publish a catalogue missing two hundred thousand stars and the runner would cache it for the weeks
after. `fetchGaiaStars` now refuses fewer than 95% of the 412 765 rows its query holds, as its
sibling query already did, and refuses an answer that fills the row limit.
`fetchText` retried the request but not the body: a connection reset part-way through the 57 MB
CSV rejected out of the loop, with no wait and no second attempt. The read now happens inside it.
Also corrected: the merge gate's account of the HYG survivors (two thirds of them are stars Gaia
measures but the main query never downloads, since Gaia puts them past the 250 pc cutoff), and the
refresh workflow's comment on what happens when the archive is unreachable.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
- The budget counted CSS pixels; lines are drawn in device pixels, so a screen scaled to 150% or
200% drew 1.5-2x the calibrated line. It now counts the canvas's drawn pixels.
- A graph was re-asked only when the drawn stars changed, so with a star budget covering the whole
catalogue, or a resize, its budget and centre stayed wherever the layer was turned on. A view that
chose its stars again now asks, and a graph is rebuilt when the stars, the range or the budget
changed (the budget by more than half the margin, or its centre by more than 5 pc).
- From inside a system the budget was worked out in astronomical units about the system's origin.
Graphs are now asked for in parsec space only; the flight back out asks.
- Comparing budgets let a request re-asked with a slightly different one supersede its twin, and the
twin's rejection cleared the state of the request that replaced it. A rejection now clears it only
for the latest request.
- The worker sorted every link to keep a few thousand, 2.2x an unbudgeted build. It now bands links
by distance, keeps every band before the one the budget runs out in, and sorts only that one:
142-168 ms on the real catalogue against 233-388 ms, 103 ms unbudgeted, returning early when all fit.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
With the drawn set following the view, the layer at 8 pc cost the integrated Radeon 503 ms a frame
at 30 pc from the Sun. Measured, the cost follows the length of line on screen (about 10 ms per
million pixels near the Sun), not the number of links: 100 000 links were 12 ms at the opening
view, 25 000 were 61 ms at 30 pc. So the budget is a length: a million pixels, turned into parsecs
at the depth the view is centred on, spent on the links nearest that centre by their nearer end.
Orbiting with links at 8 pc on the iGPU: 12-18 ms p50 at every pose measured, no long tasks.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
- Flights: the drawn stars were checked once a label pass, and a flight outruns that. Leaving a
system jumps the camera to face another way, then zooms out forty-fold in a second: 74-83% of
the stars that belong on screen were missing on the first frames back in parsec space, 33-50%
before each re-choice on the way out. While the rig animates, the check now runs every frame.
Probe on real flights (Gl 806, Barnard's Star, out): 0.90-1.00 of a fresh choice drawn on
screen in flight, 1.00 on the frame of the jump; frame p95 12.2 ms, no long tasks. Choosing for
the whole sky during flights was tried first and measured worse (0.15-0.18).
- Portrait frames: the turn and pan limits use the narrower half-extent, not the height.
- Links: a new drawn set arms the rebuild timer only when none is pending, so a continuous orbit
gets a graph at most every 250 ms instead of never; an unchanged set asks for nothing.
- The centre-move test lets the first pass happen before moving the centre.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Graph requests were never shared, on the grounds that the scene never asks for the same graph
twice. It does: turning the layer off and on while the worker is busy asks again for the graph
already waiting. The new request superseded the old one, and the old one's rejection handler,
which finds its request by range and drawn list, wiped the state of the new one: the layer stayed
on with no graph. An identical request now shares the outstanding promise, a graph being the same
when its range matches and its drawn list is the very same array.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The drawn set was two spheres, around the view's centre and around the Sun, then the brightest
stars anywhere, so most of the budget sat behind or beside the camera: at 30 pc from the Sun
15.8% of the drawn stars were on screen, at 5 pc 9.1%, in a plan view zoomed to 10 pc 3.8%.
The same tiers are now taken only from the camera's frame, widened by a quarter
(VIEW_MARGIN), with the planet hosts in view drawn first after the pinned stars, so every ring
circles a star that can be clicked. The set is chosen again at the label cadence once the view
has turned, zoomed or moved half the margin, switched projection or been resized, and once for
the whole sky on the way out to the Galaxy. Drawn stars on screen: 74-76% at 30 pc, 73-75% at
5 pc, 66% in the zoomed plan view, 91.5% at the opening view.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Same stars in the same order, for less: one walk of the brightness index, reading positions laid
out in that order, sorts the view's neighbourhood, the Sun's and the rest as it goes, instead of
gathering both neighbourhoods in catalogue order and sorting them. A refocus in the page drops
from 11.3 ms to 4.6 ms (median; worst 19.1 to 7.1). This comes before the drawn set follows the
camera's turns, which makes refocusing far more frequent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The jump-link graph linked the whole catalogue: 3.7 million links at 8 pc, 7.4-7.8 s in the
worker and a 443-515 ms frame on the main thread when they landed, and most of them between stars
that were neither drawn nor clickable. A graph request now carries the star field's drawn stars,
and the worker links only those, over an index of its own with cells as wide as the range. The
scene asks again once a new drawn set has held still for 250 ms.
The renderer is handed the graph's bounding sphere instead of computing it: three.js walked every
vertex on the main thread in the first frame that drew a new graph, 48-55 ms at 8 pc.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The panel was unmounted with its tab, so leaving it reset departure, destination and range. The
range reset was also a lie: the slider came back at 3 pc while the scene kept drawing the graph at
the range last chosen. The panel now stays mounted and is hidden while another tab is open, which
still replays the acquire wipe when it is shown again.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The adversarial review confirmed three defects in this PR, all reproduced in
the browser.
1. Superseded graphs queued up in front of routes. The worker answers
messages one at a time and cannot drop one it has started. With the
jump-link layer on, every pause on the range slider posted a full graph
build, seconds of work at 6-8 pc. Answers no longer wanted were thrown
away only once built. A route asked for afterwards waited behind every
one of them: a one-jump route took 44 s.
RoutingClient now holds requests and sends them one at a time. While one
is out, only the latest of each kind waits: a newer graph replaces an
older one before it is ever built, and the older promise is rejected with
SupersededRequest. Routes go ahead of graphs. The same question asked
again while outstanding shares the answer rather than being worked twice,
as when the layer is turned off and on during a build.
The same scenario in the browser (layer on, range stepped 5 -> 8 pc with
400 ms pauses, then Sol to Proxima): the route came back in 110 ms. The
worker was sent "links 3, links 5, route, links 8"; 6 and 7 were never
built.
2. A worker that failed left the panel stuck. With no error handling, a
worker that failed to load (a 404 on its chunk after a redeploy) or
threw left "Plotting…" and a disabled button for good, and a graph at a
range could not be asked for again.
The worker now answers an exception with a 'failed' message, which
rejects that request. A worker that fails to load or dies is abandoned,
and what it left outstanding, and everything asked afterwards, is
answered in place. The scene releases the panel when a route fails, and
forgets a graph range that was never drawn so it can be asked for again.
3. Nothing type-checked the worker. The application builder never reads
webWorkerTsConfig, and bundles the worker with esbuild, which strips
types without checking them. tsconfig.app.json leaves the file out. A
type error in the worker shipped.
`npm run worker:typecheck` (tsc -p tsconfig.worker.json) now runs in CI
beside the other project checks. webWorkerTsConfig is removed from
angular.json, since it only suggested that something checked the worker.
Tests with a fake worker cover one request at a time, a waiting graph
replaced and a route sent ahead of it, a question shared, a failure rejected
and the next request sent, and a failed worker's requests answered in place.
A scene test covers the panel released after a failed route. Negative
controls, each caught: several requests sent at once, a waiting graph kept,
graphs ahead of routes, a question asked twice, a failure answered as a
success, a failed worker waited on, the panel left pending, and a type error
in the worker (caught by worker:typecheck).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The label and star-field review fixes arrive under the routing client: the
scene keeps constructing RoutingClient beside the neighbourhood, and builds
the brightness index where it built the order. Both sides' new scene tests
are kept.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The adversarial review confirmed three costs this PR added, all reproduced in
the browser.
- The first pinned refocus stalled the first flight of a session. The
renderer built its own id-to-index Map of 423 651 entries the first time a
star was pinned, which is at the first selection, inside the approach
flight. The worst frame was 47-103 ms, and the Map stayed as a second copy
of a lookup the scene already had. The scene now pins by catalogue index,
through the StarNeighbourhood it builds at load (new `indexOf`), and the
renderer takes indices. First selection, measured in the browser: worst
frame 18 ms.
- At galactic scale every label pass rewrote the drawn set. The view centre
sweeps hundreds of parsecs a pass there, far past any star, so each pass
chose the same 70 000 stars again and uploaded 2 MB to the GPU: 11 times
on the flight out to the Galaxy. The scene no longer refocuses at galactic
scale, where the whole catalogue is a few pixels, and the renderer leaves
its buffers alone when the drawn set is unchanged. Flight to the Galaxy:
2 refocuses, no frame over 50 ms.
- At load the same set was chosen twice: once by the renderer's constructor
around the Sun, and again by the first label pass, centred on the Sun. The
scene now records the constructor's choice as the current focus.
Tests: the buffers keep their version for an unchanged set, no refocus at
load, none at galactic scale, and pins arrive as indices. Proxima's id in the
scene spec now differs from its index, so a lookup by id cannot pass for one
by index. Negative controls, each caught: an unchanged set rewritten anyway, a
refocus at galactic scale, the boot choice not recorded, and pins passed as
ids.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The label fix turns the brightness order into an index with positions and ids
laid out beside it. The star field only needs the order, so it is handed
`.order`. Both sides added scene tests in the same place; both are kept.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The adversarial review confirmed a regression in this PR. Near the Sun, the
label pass became three to four times slower than the scan and sort it
replaced.
Within about 11 pc of the Sun, and in any plan view zoomed tighter than that,
the label radius clamps to 4 pc. That sphere holds a few dozen faint dwarfs
deep in the brightness order, so the walk rarely finds fifteen stars to name
and reads nearly the whole catalogue. Reading the star objects in brightness
order jumps all over memory, so a full walk took 19-25 ms against the old
5-6 ms.
The review also found that spreadLabels checked the label count at the top of
its loop. After placing the fifteenth label it asked for a sixteenth
candidate, which near the Sun can lie at the far end of the order.
brightnessIndex now lays each star's position and id out beside the
brightness order, in that order. The walk tests stars from those arrays in
sequence and reads a star object only when it yields one. spreadLabels breaks
straight after placing the fifteenth label.
Measured on the real catalogue with the label logic reduced to what decides
placement, camera at the given distance from the Sun (old sort / this PR as
first pushed / now):
2 pc 4.9 / 24.7 / 2.5 ms
5 pc 5.7 / 23.0 / 3.1 ms
10 pc 6.3 / 18.2 / 0.95 ms
307 pc 22 / 0.01 / 0.00 ms (the opening view)
The labels are identical in every case. Now faster than the old sort at every
distance.
A new scene test counts the candidates spreadLabels takes: exactly fifteen for
fifteen labels. Negative controls, each caught: positions one axis off, ids in
catalogue order, the selected star dropped, the radius edge excluded, and the
count checked before taking a candidate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The 2026-09-14 refresh (7f187e0) regenerated exoplanets.json on main with the
host matching this stack replaces, so the file conflicted again. Resolved by
keeping this branch's, for the same reason as last time: it is the output of
the reviewed pipeline, and what main's side adds is only archive rows
published since, which the next refresh re-fetches with the fixed code.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
Route plotting ran on the main thread, and so did the jump-link graph:
- the range search for a far target, HD 2626 at 236 pc, takes 4-5 s;
- the graph at 8 pc is 3.7 million links, 6-10 s to build, then as many
link objects again to turn into vertices.
The map stopped for as long as either ran.
A Web Worker now does both. RoutingClient sends it the catalogue's ids and
positions once, and it keeps its own spatial index. A route question comes
back with the route, or with the range that would open one. A graph comes
back as one Float32Array of segment vertices, transferred rather than
copied.
On the scene side, only the latest route request is shown: an earlier
answer arriving later is dropped. Only the graph for the range last asked
for is drawn. The Routes panel says "Plotting…" and holds its button while
a request is out.
collectJumpLinks gave way to jumpLinkSegments, which writes the vertex pairs
straight into floats rather than building link objects first; the scene was
its only caller. The routing module (routing.ts) is the message protocol and
the one function answering it, so the worker is a dozen lines, and the same
answers are worked out in place where there is no Worker, as in the unit
tests' DOM. The worker is built with its own tsconfig, as the Angular
builder expects.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The star field draws a budget of the catalogue: everything within 25 pc of
the Sun, then the brightest of the rest. That choice was made once, at load,
around the Sun, and never again. On the Gaia catalogue it left most of the
map empty wherever the view went:
- a region 150 pc out drew 49 of the 442 stars within 25 pc of it;
- a plotted route ran through stars no one could see or click. Sol to
Almach at 8 pc passes 19 stars and drew 6, Sol to Mirfak 11 of 26;
- a search for a faint star flew the camera to an empty point.
The drawn set now follows the view. The scene chooses it again at the label
cadence, once the orbit target has moved more than 5 pc or the pinned stars
have changed. The budget goes, in order, to the selected star and the stars
of a plotted route, then everything within 25 pc of where the view is
centred, then the same around the Sun, then the brightest of the rest. The
instance buffers hold the budget and are rewritten in place.
Checked in Chromium on WebGPU, framing Mirfak from 12 pc: with the set
chosen around the Sun, 122 of the 649 stars within 25 pc were drawn;
following the view, all 649. At the opening view the drawn set is the same
as before.
A refocus takes 9 ms in the browser (5 ms of it choosing). The first version took
16-36 ms in the browser, a visible hitch during a flight. Most of that time
went on walking the 423 651-star brightness order once per neighbourhood,
out of catalogue order, and on recomputing 70 000 colours. Now both
neighbourhoods are gathered in one pass in catalogue order and sorted on
their own, and colours and sizes are computed once for the whole catalogue.
The brightness order itself sorts a typed copy of the magnitudes, taking
83 ms at load instead of 104-139 ms.
STAR_RENDER_BUDGET is now 70 000, and its comment gives the measurements
behind it rather than "currently set to the whole catalogue", which stopped
being true when Gaia landed. At 1920 x 1080 on a Ryzen 7700X:
- on the RTX 4080, the whole catalogue costs the same 6.1 ms a frame as the
budget;
- on the processor's two-core Radeon, standing in for an entry-level laptop,
every 100 000 stars costs about 4 ms: 112 fps at the budget, 44 at the
whole catalogue, and the same under WebGL2;
- drawn whole, the opening view turns into a grey wash that buries the
labels and the host rings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The star labels are refreshed every 0.2 s. Each pass filtered the whole
catalogue to the stars within the label radius, sorted them by magnitude,
and turned every one into a label object, all to place at most fifteen.
At the opening view the radius holds about 60 000 stars, so each pass was a
55-70 ms task on the main thread. A CPU profile of the opening view, on a
Ryzen 7700X with an RTX 4080, counted 29 tasks over 50 ms in 6.7 s, one
every 230 ms; updateLabels took 23% of the main thread. That is the stutter
the frame-time bench measured on every GPU and every render budget.
The catalogue is now sorted by brightness once, when it loads.
brightestWithin walks that order and hands stars over lazily, and
spreadLabels already stopped once it had placed fifteen labels, so a pass
reads only the stars it looks at. The output is the same as before: the
same stars, in the same order, with ties in catalogue order, the selected
star named wherever it is, and a star exactly on the radius included. The
spec checks it against the filter-then-sort it replaces. Stars are no longer
scanned at all when the view is at galactic scale, where the result was
thrown away.
Profiled again on the same view: 0 tasks over 50 ms, and the scene's
per-frame work over the window dropped from 2 028 ms to 342 ms.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The route search widened evenly from the departure, Dijkstra-style, with a
budget of 20 000 stars. On the Gaia catalogue those are all within about
40 pc of the Sun, so it found no route to anything farther at any range:
Sol to Mirfak (155 pc) failed at 3, 8, 15 and 30 pc alike. Every failure
then asked minimumRangeBetween what range would work. That search widened
the same way with a 30 pc ceiling, and it ran for up to a minute on the
main thread before giving up with nothing.
routeBetween is now an A* search. Each star is queued by the distance
travelled to it plus the straight line on to the destination, on a binary
heap rather than a linear scan of the frontier. It heads for the
destination instead of flooding the core around the departure.
minimumRangeBetween bisects the range, one routeBetween per step, because
whether a chain exists can only become truer as the range grows. Its
answer is always the longest hop of a route actually found, so a range it
names always opens one. Its ceiling is now the Routes panel's own
maximum, MAX_JUMP_RANGE_PC: a range the control cannot be set to is no
answer, and raiseTo already clamped any figure above it.
The spatial index keys its cells by one number packed from their three
indices instead of an "ix,iy,iz" string. A search visits up to 125 cells
for every star it expands, and building those strings was half of what a
route cost. forEachWithin hands neighbours over unsorted and uncollected,
which was most of the other half; within is now that, gathered and sorted.
The no-route line said nothing in the catalogue bridged the gap; it now
says no chain of jumps up to the panel's maximum reaches the star, which
is what was searched.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
The map had one way to read a distance: the Range readout, a number for how
far back the camera is. The local grid's five rings sat at 50 to 250 pc,
fixed and unlabelled. They said nothing from inside a 2 pc hop, and nothing
past 250 pc now that the Hipparcos stars Gaia places there are drawn.
A scale bar now sits under the scale rail. It shows the longest round length
(1, 2 or 5 x 10^n) that fits in 120 px, in AU inside a system and in parsecs
or kiloparsecs outside. It is measured at the depth the view is centred on,
since under perspective every depth has its own scale; under the plan view
it is exact everywhere.
The local grid's rings are now distances from the Sun, at a round step of
about a fifth of the camera's distance and out past the camera: 50 to 350 pc
from the opening view, 2 to 20 pc from twenty parsecs out. Each ring is
labelled with its distance, on the side facing what the view is centred on,
or across the far side of the grid when that is the Sun (the near side is
under the camera and out of frame). The survey edge at 250 pc stays called
out, as "Survey edge", whatever the step.
The rounding lives in one place, scale-bar.ts, shared by the bar and the
rings and tested there. Its formatter keeps three significant digits: one
digit, enough for the bar's round lengths, printed the 250 pc ring as
"300 pc".
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
HYG and Gaia were both cut at 250 pc, each on its own distance. A star
Hipparcos put at 200 pc and Gaia at 300 was kept by the first, never
downloaded from the second, and drawn at 200. That is where 83% of the
9 691 mid-magnitude HYG stars without a Gaia counterpart came from, and at
the median Hipparcos had them a third too close. The mirror case, Hipparcos
outside and Gaia inside, dropped the HYG row and left its Gaia entry
anonymous.
Gaia's own Hipparcos cross-match (hipparcos2_best_neighbour, a fixed DR3
table of 99 525 rows) gives a usable Gaia distance for 97 751 of them.
placementDistancePc keeps a star either survey puts inside the cutoff, and
draws every kept star at the better measurement, inside the cutoff or not.
57 121 HYG stars now sit at Gaia's distance. 6 833 of them are past 250 pc:
Zet Per 230 -> 259 pc, 35 Ori 137 -> 330, 44 Cnc 223 -> 613, and the
farthest, HIP 69445, at 8.7 kpc. 3 666 stars that Hipparcos put outside are
now kept, and 3 656 of them give a Gaia entry its name.
The cross-match is required rather than skipped when unreachable. Without
it, every one of those stars would move back to its Hipparcos distance, and
the published map would flip with the archive's availability. The ESA TAP
answered it with a 500 at first and in 102 s on the next try. So fetches
now retry 5xx and network failures twice, after 30 s and 120 s, in the
fetch every source goes through. The refresh job also carries the Gaia DR3
responses from run to run in the Actions cache: the release is frozen, and
a live re-fetch has already reproduced stars.bin byte for byte.
423 651 stars (+10), 61 168 HYG rows folded into Gaia entries (+3 656),
351 597 unnamed designations (-3 656). 10 886 HYG survivors and 23 unmerged
pairs under an arcsecond, both inside the merge gate's ceilings. The same
1 972 exoplanets have a host; KELT-4 A b and MWC 758 c now sit on their
named star.
The HUD's "Radius" becomes "Survey radius": 250 pc is where Gaia is
surveyed to, and no longer the edge of the map.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
With the merge gate in place, a Gaia DR3 outage no longer ships a HYG-only
catalogue: the ETL skips the unreachable source, and validateMerge then fails
on the survivor count. That is the right outcome and the wrong message: "68 000
HYG stars found no Gaia counterpart" sends the reader looking at the merge.
Gaia contributing nothing is now checked first, by name.
Two comments said Gaia was best-effort, in data-refresh.yml and on the merge
in fetchStars. They now say what happens instead.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jxMkwA2rbicdGxHosecYi
SenrokaiandClaude Opus 5.5