A nilgai—from Hindi nīl (blue) and gāy (cow)—is a large blue-grey antelope native to Asia, sharing the grasslands of India, Nepal, and Pakistan. Supposedly there is a feral herd in Texas, but we don’t talk about that.
Nilgai is also a Go library. I’m building it to look through the Milky Way’s zone of avoidance for previously undocumented high-energy sources, particularly blazar candidates. Radio, gamma-ray, X-ray, and infrared catalogues each tell part of that story.
Basically, I’m looking for little blue dots (this is a joke, I am hilarious) in the hardest place in the universe to actually look for things. It’s fine.
Nilgai converts catalogue rows into a common record: positions, positional uncertainties, brightness, and quality flags in consistent units. That gives a caller something useful to compare across instruments. Matching sources and deciding whether a candidate is an AGN remain separate work.
| Catalogue | Coverage | Reader |
|---|---|---|
| Chandra Source Catalog | X-rays | read.Chandra, TAP text export |
| AllWISE | Mid-infrared, W1–W4: 3.4, 4.6, 12, and 22 µm | read.AllWISE, IRSA CSV export |
| Fermi-LAT 4FGL | Gamma rays | read.Fermi, FITS point-source table |
| MeerKAT | Radio, 1.28 GHz | Interface defined; reader planned |
| Neowise | Repeated mid-infrared observations at 3.4 and 4.6 µm | Interface defined; reader planned |
Readers yield a source or a *nilgai.RowError for each row. A row error
names the input, data row, and column, then reading continues. An error
that prevents reading the input is yielded once and ends the sequence.
Readers consume the supplied stream; callers open the files.
AllWISE requires W1 and can also read W2–W4. Optional bands need both magnitude and uncertainty columns. A populated magnitude with an empty uncertainty is an upper limit; an optional band with both cells empty has no measurement and is omitted.
Sourceholds the catalogue’s identifier, an ICRS position in degrees, an error ellipse, the position epoch when known, and measurements.ErrorEllipsestores one-sigma semi-axes in arcseconds and a position angle in degrees east of north. It also retains the catalogue’s stated axes and confidence level so the conversion can be checked.Measurementrecords a detection or upper limit through a named filter, such as W1, and the libtheme bands it covers.FluxDensityandEnergyFluxare held in SI;JanskyandErgPerSecPerCm2make them from the units catalogues give. A value and its uncertainty are one type. Original value text and catalogue flags can be retained alongside Nilgai’s quality assessment.Whenstores an MJD or an unknown time.LightCurvegroups measurements at one position in time order, ready for the planned Neowise reader.
RadiusFactor and ToOneSigma convert positional confidence levels.
read.WiseFlux converts WISE Vega magnitudes using zero points cited in
the WISE Explanatory Supplement, IV.4.h, after Jarrett et al. 2011,
ApJ 735, 112. Its symmetric flux uncertainty is a first-order approximation.
Units and spectral conversions come from
libtheme. Wavelengths are in
nanometres. band.BandsBetween maps a filter’s bounds onto libtheme’s
bands.Band; Nilgai has no separate spectral band type. Fermi’s configured
colour profile spans 50 MeV to 1 TeV; the Energy_Flux100 measurement
filter covers 100 MeV to 100 GeV.
The root package defines the record. Three packages build on it:
bandholds filter edges, citations, and false-colour profiles.readparses catalogues and converts their values.skydraws one-sigma error ellipses and arranges panels withStrip.
read.Unassociated selects Fermi sources with a blank CLASS1 within an
exclusive latitude bound around the Galactic plane. Sources classed unk
or bcu are excluded by this selection. Errors pass through unchanged.
The command uses puzzlemouth.
nilgai show draws selected searches around a position as a PNG:
nilgai show --chandra chandra.tsv --allwise allwise.csv \
--ra 266.4168371 --dec -29.0078106 --width 12 --pixels 600 \
> sgra.pngAdd --fermi catalogue.fit to include 4FGL. A width list such as
--width 12,6 draws panels side by side. Each source has its instrument’s
colour, adjusted to the reader’s contrast requirements through libreadme.
Unreadable rows are reported on standard error after the image is written;
the image still contains the valid sources, and the command exits with an error.
nilgai orbits runs the fastest stars of a Gaia product back through
the Milky Way to their nearest pass by Sgr A*, on the GPU:
nilgai orbits --gaia ~/sky/gaia/gaia_source_0_0_r180.fits --sun frame \
> candidates.tsvThe product is what celestial-data's fetch --archive gaia brings
home. The stars whose parallax is at least --quality times its error
are kept, the fastest --fastest of them by their speed against the
Sun are selected, and those run back --span years in --steps steps,
--batch at a time, through celestial's MWPotential2014 with Sgr A* at
its centre. One line a star comes out, sorted by the nearest pass, under
a header that says what was kept, selected and left out. --sun is a
choice with no default: frame runs celestial's galactocentric frame as
published, model gives the frame's Sun the model's speed; the two
differ by 20 km/s along the rotation. The numbers the flags default to
are in cmd/nilgai/orbits.yaml, each with its reason. The command
builds on macOS only now, since the kernel is celestial's Metal one.
searches/ holds the questions nilgai asks of the sky, one folder each.
The first is firang: stars thrown out of
the galactic centre, read for the blanets they may have eaten there.
Use the Go version in go.mod, then run:
game build
game checkFilter edges, colour mappings, and drawing styles live in JSON data files beside their packages. Measured bounds carry citations; chosen display colours remain identified as choices. Tests include small catalogue fixtures, real exported rows, and property checks for ellipse geometry.
Jane Michelle Arc · jmich@fastmail.nl