Piazzi 0.2.1

Open source astrometry and photometry of asteroids and comets. Product page: https://datamonkeys.ai/piazzi/

Downloads

File	Platform
Piazzi-mac.dmg	macOS 10.13+, Intel and Apple Silicon. Signed and notarized by Apple.
Piazzi-windows-x64.zip	Windows 10 1803+ 64-bit, OpenGL 3.0+. Not yet tested on real Windows hardware: please open an issue if you try it.
Piazzi-linux-x86_64.AppImage	Ubuntu 22.04+ (glibc ≥ 2.35), needs libfuse2.

Checksums: SHA256SUMS.txt. The built-in example ((5157) Hindemith, three ZTF frames) runs offline; MPC submission is disabled for it.

Known issues

• The Windows build is cross-compiled and packaged but not yet tested on real Windows hardware.

Changed

• Settings format (breaking): every setting key now has its own snake_case name, for example
  focal_length_mm, pixel_width_mm, flip_horizontal, catalog_faint_mag, plate_order and trail_detect. Older
  key names are no longer read. Update existing .cfg files and saved profiles (the bundled example already uses the
  new names).
• Code comments, log messages, CLI help, CI and benchmark reports are in English.
• Added CITATION.cff, a README screenshot and badges. tools/benchmark/results/ keeps only the final runs (see its
  README).
• Tests no longer need any external download: image tests use the bundled ZTF example ((5157) Hindemith,
  cpp/gui/assets/example/) with a new JPL Horizons reference (cpp/tests/data/horizons_hindemith.txt, station I41,
  query in the header). Limits are ~2–3× the measured accuracy (O−C max 0.021″, rms 0.016″ → ≤ 0.06″ / 0.045″; plate
  rms ≤ 0.053″ → < 0.15″ with ≥ 50 references; rate/PA vs Horizons within 0.02″/min and 3°; magnitude spread
  0.09 → ≤ 0.25). test_orbits uses versioned osculating elements from Horizons in MPCORB/COMET layout
  (cpp/tests/data/orbits_*.dat, origin in cpp/tests/data/README.md); test_image_io adds a synthetic
  uncompressed SBIG file (pixels, exposure, mid-exposure time).
• Tests no longer download third-party sample data: the external download script, its CI step and the old reference
  files were removed. DESIGN.md "Accuracy" now describes the bundled ZTF example.
• README/THIRD_PARTY_NOTICES describe .cfg support neutrally (INI-style configuration files).
• run.py writes paths relative to the repository root (or just the file name when outside it) in results and
  messages (run, inject, vs-mpc, check-ades); existing results under tools/benchmark/results/ were rewritten
  without local absolute paths (<local>/…).
• Core error messages in the C API are now English and actionable, with a stable code (no_images,
  image_unreadable, catalog_empty, internal) and optional detail in the error JSON; the apps translate by code.
• astro_reduce honours plate_ruwe like the CLI (default 1.4, unchanged behaviour).

Fixed

• Without --object/object name the API no longer labels every mover "Piazzi": the designation is empty, ADES carries
  only the trkSub and the MPC report uses the per-track temporary designation; apps show the trkSub or "objeto N".
• Result JSON never contains NaN/Inf (written as null); added per measurement trksub, sig_ra, sig_dec (″, the
  ADES σ), trail_len (″), remarks, and per mover trksub (existing fields unchanged).
• Undefined behaviour in the linking grid hash (left shift of a negative value).
• Shift-and-stack search: forced measurements with non-positive flux are discarded (no log10 of ≤ 0).
• astro_image_catalog: trail fragments get type T and attached trails L (R stays "reference").
• Dockerfile header no longer claims that the image runs the tests.
• ADES # telescope no longer written with an empty design and aperture 0.0 when telescope_design/telescope_aperture_m are
  unset: both are derived from the telescope description (all ZTF/SDSS/NEA benchmark reports were affected).
• Trail centroid covariance (ADES σ of trailed movers): Poisson-weighted sandwich covariance of the unweighted trail fit
  instead of a uniform χ²/(n−p) scale (trail_cov_poisson, default 1); SDSS trail injection |O−C|/σ_c 0.795 → 0.706.
  run.py inject adds per-axis ratios against σ_c (the injection O−C has no plate error by construction).
• 80-column report: a seconds field rounding to 60 no longer prints "-0.0" (e.g. -00 58 -0.0), which broke the Dec
  columns 45–56 (2 SDSS measurements); same clamp in RA.

Added

• astro_settings_schema(): JSON list of the exposed settings (key, group camera/detection/plate/motion/reports/
  advanced, English and Portuguese label and hint, type number/int/bool/enum/text with options, min/max, core default),
  including the new gain, trail_fit (−1 auto, 0 off, 1 trail, 2 star PSF), trail_detect, stack_search,
  plate_weights, plate_color, motion_sigma and plate_ruwe. Desktop, iOS and web settings screens are generated from
  it: collapsible groups ("Advanced" closed), checkboxes, menus, range-checked numeric fields, hints, restore defaults.
• English user interface by default with optional Portuguese in the three apps (desktop/web remember the choice; iOS
  follows the system language with an in-app override). Strings centralised per app.
• "Open example": bundled offline example (three 6′ ZTF crops of (5157) Hindemith, 2021-02-06, Gaia DR3 excerpt,
  example.cfg, ~1.7 MB) in cpp/gui/assets/example/, web/example/ and the iOS bundle; MPC submission is blocked
  for example data and its observatory code is XXX. Credits (ZTF/IRSA, ESA/Gaia/DPAC) in the example README and
  THIRD_PARTY_NOTICES.
• App icon (packaging/icon/piazzi.svg, PNG 16–1024, .icns, .ico, Windows .rc, iOS AppIcon, macOS bundle icon,
  packaging/icon.png for the AppImage, SDL window icon).
• run.py check-ades: offline ADES 2022 (PSV) and MPC 80-column format checker for the benchmark outputs or single
  files; ctest report_format applies it to the reports of the bundled example.
• astro_version(), piazzi --version, single version source (cpp/include/piazzi/version.h) used by CMake and
  shown in the desktop, iOS and web home screens.