pyredshift 1.9

Interactive redshifting of 1D astronomical spectra (Python successor of pdlredshift / redshift.f)

pyredshift

Interactive redshifting of 1D astronomical spectra, by eye.

pyredshift plots a spectrum histogram-style with a full-window crosshair cursor and overlays rest-frame line identifications that follow a trial redshift. Mark a feature you recognise, and the line labels jump to that redshift; refine, zoom, bin, smooth, fit a continuum and measure equivalent widths — all from single keypresses. It is the Python descendant of the author's venerable redshift.f (Fortran/Figaro, 1990s) via pdlredshift (Perl/PDL/PGPLOT), and deliberately keeps that fast keyboard-driven feel.

An SDSS quasar from Example-Spectra at z = 2.48 and counting, in --retro mode: Lyα, CIV and CIII] lined up; the crosshair cursor mid-hunt.

Install

pip install git+https://github.com/karlglazebrook/pyredshift

or from a checkout:

pip install .

This installs the pyredshift command into the active python's bin/ (e.g. anaconda3/bin) and the pyredshift package into site-packages. Requires python ≥ 3.9 with numpy, matplotlib, astropy and scipy.

Usage

pyredshift spectrum.fits              # format auto-detected
pyredshift spectrum.fits -f sdss      # force a format
pyredshift spectrum.fits -z 2.19      # start from a known redshift
pyredshift spectrum.fits --retro      # retro PGPLOT-style black background
pyredshift spectrum.fits --microns    # keep micron wavelengths (no A conversion)

The file format is auto-detected by inspecting the FITS structure and column names/units:

Format	Description
fits	1D (or 2D — row 1) FITS image with WCS; CD1_1 overrides CDELT1, IRAF/SDSS log-λ flags handled
sdss	SDSS binary table (LOGLAM/FLUX, either case)
table	generic binary table with wave+flux columns; TUNIT (µm, nm, Å) converted automatically (aliases: jwst, gabe, dja)
xs	XSHOOTER 1D table spectrum (nm)
xs2	XSHOOTER as a FITS image in extension 1
outthere	OutThere multi-extension grism spectra, stitched and flat-calibrated
csv	comma-separated, header rows skipped
ascii	2-column text

Wavelengths that look like microns are converted to Ångstroms; NaNs are treated as bad pixels and plot as gaps, with orange markers along the bottom of the frame.

Real spectra to practice on are included in Example-Spectra — the true redshifts are hidden in their FITS headers.

From a Jupyter notebook

The interactive module also works straight from a notebook, given numpy arrays:

from pyredshift import redshift as redshift_module
z = redshift_module.redshift(wave, flux)

The interactive window opens outside the notebook (inline backends cannot deliver events to it) and the cell blocks until you quit with q; the final annotated view is then embedded in the cell output and the redshift returned. Any non-interactive backend (e.g. %matplotlib inline) is switched to a GUI backend for the session and restored afterwards — including if you interrupt the kernel mid-session.

Local kernels only: the window opens on the machine where the kernel runs, so remote kernels (JupyterHub, cloud, ssh without X forwarding) cannot work — pyredshift detects this and says so rather than hanging or crashing the kernel.

Interaction

Press ? (or click the ? button) for the full command reference — it opens as a themed page in your browser, with the complete line list and a short guide in separate tabs. The essentials:

• Identify a line: put the crosshair on a feature and press ESC followed by a shortcut key (a = Hα, b = Hβ, l = Lyα, o = [OII], c = CIV ...) — or right-click the feature and pick the line from the popup menu, or g to type any rest wavelength. Press ESC ESC on a sharp line to refine the redshift; = enters a redshift directly.
• Navigate: drag the mouse for a rubber-band zoom (a purely horizontal drag zooms X only), z/u and i/o zoom about the cursor, [ ] pan, w whole range, a autoscale, h home. The matplotlib toolbar works too and stays in sync.
• Analyse: b bin, s smooth, _ iterative continuum fit, m equivalent width and line flux, B zap artefacts, p print to PDF.
• Quit: q — the final redshift is reported on the terminal.

Line identifications come from pyredshift.lines (CSV: wavelength, name, label, matplotlib colour, comment — all vacuum wavelengths; a colour may be a light/dark pair for the two backgrounds), installed alongside the module and easy to customise. The window size and aspect ratio are remembered between runs in ~/.pyredshift.json. A template spectrum can be overlaid with t (searched for in $DATADIR, $NEWMODELS/Spectra, then ~/Templates/Spectra).

Layout

pyredshift            the command-line script: format detection and reading
src/pyredshift/       the interactive module (redshift.py) + data files

The script prefers a src/pyredshift/ found next to it or under the current directory, so a development checkout always runs its own code.

Credits

Karl Glazebrook, Swinburne University of Technology.

Lineage: redshift.f (FIGARO/Fortran, ~1992) → pdlredshift (Perl/PDL) → pyredshift. Ported with the assistance of Claude (Anthropic).

MIT license — see LICENSE.