Change History#

V1.3.8 Not Released#

  • Documentation update

  • Fix DCR bug that was present when called from CLI.

  • Fixed Bug when using relative path for selecting master flat

  • Fixed Bug when dcr didn’t processed properly when comparison lamps have OBSTYPE = ARC

V1.3.7 17-08-2023#

  • Updated documentation theme to allow dark mode

  • Removed python 3.6 and 3.8 from testing and added 3.9 and 3.10

  • Unfixed numpy version

  • Removed testing with conda on github action because it only added complexity

  • Added 3 seconds timeout to request to github to validate version and check if running version is latest

  • Added correct exception handler in case the http request timed out.

V1.3.6 25-03-2022#

  • Updated HgArNe reference lamps for 400M1 and 600MID GG395

  • Fixed deprecation warning for numpy.float conversion

V1.3.5 24-11-2021#

  • Created –target-min-width and –target-max-width flags to allow extraction of extended sources.

  • Replaced coveralls with codecov for code coverage.

  • Documentation improvements and updates.

  • Fixed issue with extracted spectrum that where masked in 2D.

  • Code style improvements.

V1.3.4 11-05-2021#

  • Updated default search pattern for redspec

  • Fixed branch name in Github Actions

V1.3.3 19-04-2021#

  • Fixed bug that caused overscan and bias correction happen in the same image resulting in double bias level subtraction.

V1.3.2 09-10-2020#

  • Fixed Github Actions setup

  • Removed pandas version constraint and implemented workaround to be able to use latest pandas version.

  • Modified installation of dcr on travis and Github Actions

V1.3.1 23-09-2020#

  • Documentation improvements:

  • Added more test code.

  • Cleaned .travis.yml and created special dcr installation script for travis.

  • Changed the way core.combine_data names new files.

  • Fixed version checker due to deprecation of access token as url parameter.

  • New core.identify_technique. Was developed in the web application context.

  • Created –skip-slit-trim argument to provide more control for certain use cases.

  • Removed python 3.5 because it will not be supported anymore.

  • Improved AEON Support:

    • Added values for OBSTYPE and required logic.

  • Bugs Fixed:

    • Serial and Parallel binning extraction from header was not working

    • Changed url for astroplan’s server

V1.3.0 06-03-2020#

  • Made it compatible with Astropy 4.0

  • All versions are free except for Pandas [#314]

  • wavelength.WavelengthCalibration.__call__ can now return a json output.

  • core.setup_logging can now create a generic logger (same format).

  • Modified how master bias are named.

  • Removed bias overscan and trimming correction on master bias creation.

  • Bugs Fixed:

    • –max-targets was not being used, missed connection in MainApp.

  • Updated keyword values of all reference lamps in the library according to [#292]

  • Refactored wavelength.WavelengthCalibration class moving several methods to core [#300, #303]

  • Refactored wavelength.WavelengthCalibration to be instantiated without arguments.

  • Improved messages at critical stages of wavelength calibration.

  • Moved setup_logging call from main package __init__ to scripts or entry points, this allows to re use other master loggers.

  • Changed –background-threshold to multiply by detection limit instead of background level

  • Created standard JSON output for WavelengthCalibration.

V1.2.1 19-08-2019#

  • Bugs fixed

    • Bias process was not fully ignored when –ignore-bias was used [#289].

    • pandas version was not specified in environment.yml [#288, #290]

    • Target extraction failed for low signal targets because background subtraction was being ignored at the step of actually identifying targets.

  • Install instructions updated [#290]

  • Moved static methods from ImageProcessor to core.

  • Added function to validate ccd regions using regular expressions.

  • Using lamps keywords to select reference lamps.

  • Replaced target_stddev by target_fwhm in function extract and extract_fractional.

  • Replaced nsigmas by nfwhm everywhere.

  • Added argument –background-threshold with default value 3.

  • Added argument –fit-targets-with with options moffat and gaussian.

V1.2.0 26-10-2018#

  • Bugs removed:

    • If there was more than one lamp for a science target the lamp recorded as used was one of them only.

    • A percentage symbols was added to the help of --saturation argument, this caused a crash when redccd -h or redccd --help was used.

  • Numpy is fixed to the version 1.15.2 until further notice.

  • Reference lamps now get the extraction window added to the end of the file name. This is to avoid overwritting the lamps when they were used for more than one target.

  • DCR install script is now more advanced and requires a virtual environment to work on.

  • Added SOAR Logo to ReadTheDocs page.

  • Changed install instruction with exact steps and commands instead of referencing documentation.

  • Improved method to detect saturated images. Added a table with the half full well for all the readout modes possible and created a method to easily retrieve the value. This is a big improvement since in earlier versions the saturation limit was set to 65000 ADU regardless the input data and the user had to set a different one using the argument --saturation.

  • Repurposed the --saturation command line argument, now is used to define the percentage of pixels above the saturation level, which for simplicity is the value of half full well. A default value of 1 percent was set as default.

  • Added record information of target trace into the header and logs.

  • Added record of background extraction regions into the header and logs.

  • Made all plots full screen and the images using the gray cmap.

  • Trace information is printed in the logs and also is recorded in the image’s header

  • Added sigma clipping to target tracing functions

V1.1.2 05-10-2018#

  • Version 1.1.2 is pip instalable

    pip install goodman-pipeline

  • Project and package renamed to goodman_pipeline this is because the previous was too generic. Now we have this structure:

    goodman_pipeline/
       docs/
       goodman_pipeline/
          core/
          images/
          ..etc
       setup.py
       ..etc
    
  • Bugs Fixed:

    • DataFrame index is unusable when partial parts are eliminated. Added index_reset(drop=True)

    • Data conversion from string to integer needed to be converted to float first.

    • For low SNR data there was confusion of noise with targets, added a median filter and increased the the order value of peak detection.

  • Created several new keywords:

    GSP_EXTR:

    Extraction window at the first column.

    GSP_SCTR:

    Used for extracted comparison lamps, contains the name of the file of science target that the lamp was extracted for.

    GSP_LAMP:

    For science targets, it records the name of the lamp used for the wavelength calibration.

  • “Sliding” cross correlation window (to trace non-linearity of wavelength solution) is set to the maximum value between the length of the lamp spectrum in pixels and four times the global cross correlation of the reference lamp to the new one.

  • Iterations in sigma clipping of differences between obtained wavelength values and laboratory values was increased from 1 to 3. This is for removing bad fitted lines and also RMS error calculation.

  • Gaussian Kernel size for reference lamp convolution is now dependent on slit size and binning

  • Added reference lamps for all gratings and their modes except 1200M0

  • Created script install_dcr.sh

  • Increased code coverage

  • Eliminated None elements in list of instances of goodman_pipeline.core.core.NightDataContainer

  • Improved several logging messages

    • In general, it informs more, when it does an action and when it does not. What files are discarded,

    • Debugging plots are more complete for identify_targets.

  • Created new argument --debug-plot dedicated for graphical debugging, the old --debug will show additional messages but will not produce any graphical output.

  • Removed ability to process several folders in sequence, now the pipeline has to be run for each folder separately.

V1.1.1 23-08-2018#

  • Bugs Fixed:

    • Added clean exit when pipeline is unable to determine instrument or technique used.

    • Conversion from string to integer not always works, added intermediate float conversion.

    • Abrupt exit when there were non-fits-compliant keywords. Now it attempts to fix them all automatically and warns the user. Also, it ends the execution and informs the user to try again.

  • Removed unused code and tools.

  • Relocated module goodman_pipeline.core.check_version to pipeline/core.

  • Implemented Authorized GitHub API access and added actual version check

  • Moved command line interface from goodman/bin/ to goodman/pipeline/script/

  • Specified version of cython to be able to build.

  • Added reference lamps for all usable modes for the grating 600 l/mm

  • Created method to use automatic keyword fix from ccdproc.

  • Improved help information of arguments

  • Documentation updates

V1.1.0 24-07-2018#

  • Bugs fixed

    • --keep-cosmic-file would work for dcr but not for lacosmic

  • Changed organization of ReadTheDocs information

    • New structure

    • Added references to external packages

    • This page is the single place to add changes information. CHANGES.md still exist but contains a link here.

  • Added --version argument.

  • Implemented astroscrappy’s LACosmic method

  • removed ccdproc’s cosmicray_lacosmic().

  • created default method for cosmic ray rejection.

    • For binning 1x1 default is dcr

    • For binning 2x2 default is lacosmic

    • For binning 3x3 default is lacosmic

methods dcr, lacosmic or none can still be forced by using --cosmic <method>

V1.0.3 11-07-2018#

  • Bugs fixed

    • programatically access to the version number did not work because it was based purely on setup.cfg now setup.py has a function that creates the file pipeline.version which is accessed by pipeline/__init__.py

    • File naming was making some file dissapear by being overwritten for files that contained more than one target the next file name would match the previous one. A differentiator was added.

V1.0.2 10-07-2018#

  • Removed module goodman/pipeline/info.py and placed all metadata in goodman/setup.cfg.

  • Several updates to documentation

    • Added comment on how to organize data on soardata3.

    • Added link to licence on footer.

    • User manual now is in ReadTheDocs and no longer available as a pdf.

    • Improved information on debug plots

  • Bugs Fixed.

    • fixed GSP_FNAM value for reference lamps

    • Spectral limit calculation by including binning into the equation

    • Included binning in the calculation of the wavelength solution

    • Corrected messages and conditions under which the prefix for cosmic ray rejection is used

    • Image combination call and messages

  • Other additions + Added lookup table dcr.par file generator and found optimal parameters for Red camera and binning 2x2

V1.0.1 xx-xx-2018#

  • Moved user manual from external repo to goodman/docs/

  • Added version checker

  • Centralised metadata (__version__, __licence__, etc) in goodman/setup.cfg

  • Added CHANGES.md

V1.0.0 29-04-2018#

  • First production ready release