4. The reduce command

4.1. Introduction

The reduce command is the DRAGONS Recipe System command line interface. The Recipe System also provides an application programming interface (API), whereby users and developers can programmatically invoke Reduce and set parameters on an instance of that class (see The Reduce Class).

Both interfaces allow users to configure and launch a Recipe System processing pipeline on one or more similar input datasets. Control of the Recipe System on the reduce command line is provided by a variety of options and switches which we will introduce in this chapter.

4.2. Usage Examples

Below we show examples that a user might typically want to do when using reduce. The command offers a lot of flexibility though, these examples are just a small subset of the possibilities. The objective here is to help the user get started.

4.2.1. Nominal usage

Because the Recipe System is automated, in many cases all that is needed is the command and a filename.

reduce S20161025S0111.fits

The system defaults to the “sq” mode, ie. science quality recipes. The best match recipe will be used with the best match primitive set. The required processed calibrations will be fetched from the local calibration manager.

The system defaults to using the Gemini Astrodata configuration package and the Gemini data reduction package, gemini_instruments and geminidr, respectively.

4.2.2. Overriding Primitive Parameters

The primitives for each set are given default values that have been found to give good results in most cases. Depending on the data and the science objectives, it might be necessary to tweak the primitive parameters to optimize the reduction. The -p, or in long form --param option allows the user to override the defaults.

reduce S20161025S0111.fits -p stackFrames:operation=median \
       stackFrames:reject_method=minmax

This sets the stackFrames input parameters operation and reject_method to median and minmax, respectively.

As one can see that, if several parameters are to be modified, the command can grow rather long. There is a way to keep it clean, see the section below on the @file facility.

4.2.3. Calling Specific Recipes and Primitives

The Recipe System’s default behavior is to select the best recipe automatically. It is however possible, and sometimes required, to override this.

4.2.3.1. Override the default recipe

The first case where the recipe selection can be overridden is to select a recipe in the library different from the default. A recipe library can contain more than one recipe. Only one is set as the default. To let the Recipe System select the most appropriate recipe library, but then request the use of recipe within that library other than the default, simply state the name of the desired recipe. A good example is when making a bad pixel mask (BPM) for NIRI:

reduce @flats @darks -r makeProcessedBPM

Here the Recipes System will find the recipe library for NIRI flats (because the flats are first in the list), and then instead of running the default recipe which would in this case make a processed flat, it will run the makeProcessedBPM recipe.

For information about the @ format, see The @file Facility below.

4.2.3.2. User recipe

It is possible for the user to force the use of a custom recipe. This is done with the -r flag again. The structure “recipe library containing recipes” must still be obeyed. Here is how the request is made:

reduce S20161025S0111.fits -r myrecipelibrary.myspecialrecipe

Both the name of the recipe library and, after the dot, the name of the recipe function are required. The path to the library can be prepended.

4.2.3.3. Calling a single primitive

Single primitives can be called directly from the command line bypassing the recipes entirely. A useful case is when one wants to display dataset. There is a primitive named display. The Recipe System will find the best-match primitive set, and then run the display primitive it contains.

reduce S20161025S0111.fits -r display

4.2.4. Manually Setting Calibrations

When the calibration manager is not available or if working on a new type of data not yet coded in the calibration association rules, it will be necessary to specify the processed calibration to use on the command line.

Another situation would be if one wanted to try various version of a calibration or different calibrations altogether to try to optimize a reduction. In such a case, one needs full control on which calibration is being used rather than always using the “best-match” returned by the local calibration manager.

reduce S20161025S0111.fits --user_cal processed_bias:S20161025S0200_bias.fits

4.3. Command Line Options and Switches

The reduce command help is provided by the --help option. This help is also available as a manual page as (man reduce). The options and switches are described further here.

4.3.1. Information Switches

-h, –help
show the help message and exit
-v, –version
show program’s version number and exit
-d, –displayflags

Display all parsed option flags and exit.

The table provides a convenient view of all passed and default values for reduce. This can be useful when wanting to verify the syntax of a reduce call and to make sure everything has been parsed as expected.

Note that when not specified, recipename indicates ‘None’ because at this point in the execution the Recipe System has not yet been invoked and a default recipe not yet been determined. Eg.,

$ reduce -d --logmode quiet fitsfile.fits

   Literals                      var 'dest'              Value
  -----------------------------------------------------------------
 ['-d', '--displayflags']        :: displayflags         :: True
 ['-p', '--param']               :: userparam            :: None
 ['--logmode']                   :: logmode              :: quiet
 ['--ql']                        :: mode                 :: sq
 ['--qa']                        :: mode                 :: sq
 ['--upload']                    :: upload               :: None
 ['-r', '--recipe']              :: recipename           :: None
 ['--adpkg']                     :: adpkg                :: None
 ['--suffix']                    :: suffix               :: None
 ['--drpkg']                     :: drpkg                :: geminidr
 ['--user_cal']                  :: user_cal             :: None
 ['--logfile']                   :: logfile              :: reduce.log
 -----------------------------------------------------------------

Input fits file(s):      fitsfile.fits

4.3.2. Configuration Switches and Options

–adpkg <ADPKG>

Specify an external AstroData configuration package. This is used for non-Gemini instruments or during development of a new Gemini instrument. The package must be importable. The default AstroData configuration package is gemini_instruments and it is distributed with DRAGONS.

E.g., --adpkg scorpio_instruments

–drpkg DRPKG

Specify an external data reduction package. This is used for non-Gemini instruments or during development of a new Gemini instrument. The package must be importable. The default data reduction package is geminidr and it is distributed with DRAGONS.

E.g., --drpkg scorpiodr

–logfile <LOGFILE>
Set the log file name. The default is reduce.log and it is written in the current directory.
–logmode <LOGMODE>

Set logging mode. One of

  • standard
  • quiet
  • debug

“quiet” writes only to the log file. The other modes writes information to the screen and to the log file. The default is “standard”.

-p <USERPARAM [USERPARAM …]>, –param <USERPARAM [USERPARAM …]>

Set a primitive input parameter from the command line. The form is

-p primitivename:parametername=value

This sets the parameter such that it applies only for the primitive “primitivename”. To set multiple parameter-value pairs, separate them with whitespace, eg. -p par1=val1 par2=val2

The form -p parametername=value is also allowed but beware, that will sets any parameter with that name from any primitives to that value. It is somewhat dangerous and of limited use. It is to be seen as a global setting.

–qa
Set the mode of operation to “qa”, “quality assessment”. When no “qa” or “ql” flag are specified the default mode is “sq”. The “qa” mode is use internally at Gemini. Recipes differ depending on the mode.
–ql
Set the mode of operation to “ql”, “quicklook”. When no “qa” or “ql” flag are specified the default mode is “sq”. The “ql” mode is use for quick, near science quality reduction. Science quality is not guaranteed. Recipes differ depending on the mode. This mode is not yet implemented. “ql” recipes are not yet available.
-r <RECIPENAME>, –recipe <RECIPENAME>

Specify a recipe by name. Users can request a non-default system recipe by names, e.g., -r makeProcessedBPM, or may specify their own recipe library and recipe function within. A user-defined recipe function must be “dotted” with the recipe file.

-r /path/to/recipes/recipelibrary.recipename

For a recipe file in the current working directory, the path can be omitted:

-r recipelibrary.recipename

A recipe library can contain more than one recipe. The recipe library must be a Python module, eg. recipelibrary.py. The recipes are Python functions within that module.

Finally, instead of specifying a recipe, it is possible to specify a primitive:

-r display
–suffix <SUFFIX>
Add “suffix” to output filenames at the end of the reduction.
–upload

Currently used internally (Gemini) only.

Send specific pipeline products to internal database. The default is None.

--upload metrics calibs

or equivalently:

--upload=metrics,calibs
–user_cal <USER_CAL [USER_CAL …]>

Specify which processed calibration to use for the reduction. This override the selection from the local calibration manager. The syntax is:

--user_cal calibrationtype:path/calibrationfilename

Eg.:

--user_cal processed_bias:somepath/processed_bias.fits

The recognized calibration types are currently:

  • processed_arc
  • processed_bias
  • processed_dark
  • processed_flat
  • processed_fringe
  • processed_standard

4.4. The @file Facility

The reduce command line interface supports an “at-file” facility. An @file allows users to provide any and all command line options and flags to reduce in an acsii text file. This tool is very useful to keep the command line to a reasonable length and also to keep a record of the configurations that are applied. Here we illustrate how to use it.

4.4.1. Basic @file Usage

In a previous section we had an example where we were modifying a primmitive’s input parameter values.

reduce S20161025S0111.fits -p stackFrames:operation=median \
       stackFrames:reject_method=minmax

Instead of typing the parameter settings on the command line, it might be more convenient to use an “at-file”. We can write the parameter information in the “at-file” and add it to our reduce call. Let us have a file named “myreduction.par” with this content:

-p
stackFrames:operation=median
stackFrames:reject_method=minmax

Now we can call reduce as follow:

reduce S20161025S0111.fits @myreduction.par

By passing an @file to reduce on the command line, users can encapsulate all the options and positional arguments they may wish to specify in a single @file. It is possible to use multiple @file and even to embed one or more @file in another (see Recursive @file Usage). The parser opens all files sequentially and parses all arguments in the same manner as if they were specified on the command line.

To further illustrate the convenience provided by an @file, we’ll continue with an example reduce command line that has even more arguments. We will also include new positional arguments, i.e., file names:

$ reduce -p stackFrames:operation=median stackFrames:reject_method=minmax \
  -r myrecipelib.myrecipe S20161025S0200.fits S20161025S0201.fits \
  S20161025S0202.fits S20161025S0203.fits S20161025S0204.fits

Here, two user parameters are being specified with -p, a recipe with -r, and a list of input datasets. We can write all this into a plain text @file, let’s name it “reduce_args.par”:

# input data files
S20161025S0200.fits
S20161025S0201.fits
S20161025S0202.fits
S20161025S0203.fits
S20161025S0204.fits

# primitive parameters optimization
--param

    # stackFrames
    stackFrames:operation=median
    stackFrames:reject_method=minmax

# recipe
-r
    myrecipelib.myrecipe

Now we can call reduce this way:

reduce @reduce_args.par

The order of the arguments in an @file is irrelevant, as is the file name. Also, the parser sees no difference across white space characters, such as space, tabs, newlines, etc. Comments are accommodated, both full line and in-line with the # character.

Finally, the “at-file” does not need to be in the current directory. A path can be given. For example:

reduce @../reduce_args.par

4.4.2. Recursive @file Usage

As implemented, the @file facility will recursively handle and process other @file specifications that appear in a @file or on the command line. For example, we may have another file containing a list of input files, let’s call it “bias.lis”:

# raw biases
S20161025S0200.fits
S20161025S0201.fits
S20161025S0202.fits
S20161025S0203.fits
S20161025S0204.fits

Then, we can add this list as an “at-file” in the reduce_args.par file:

# input files
@bias.lis

# primitive parameters optimization
--param

    # stackFrames
    stackFrames:operation=median
    stackFrames:reject_method=minmax

# recipe
-r
    myrecipelib.myrecipe

The reduce call becomes:

reduce @reduce_args.par

The parser will open and read the @bias.lis, consuming those lines in the same way as any other command line arguments. Indeed, such a file need not only contain fits files (positional arguments), but other arguments as well. This is recursive. That is, the @fitsfiles can contain other “at-files”, which can contain other “at-files”, which can contain …, etc. These will be processed serially.

Or one might want to keep the input files and the parameter settings separate. Then if we remove the @bias.lis from the “reduce_args.par” files, we can use it explicitly on the reduce command line:

reduce @bias.lis @reduce_args.par

4.4.3. Overriding @file Values

The reduce application employs a customized command line parser such that the command line option given in the @file can be modified on the command line after the @file has been processed.

The -p or --param will accumulate a set of parameters or override a particular parameter. This may be seen when a parameter is specified in a user @file and then specified on the command line. See Example 1 and 2 below.

For unitary value arguments, the command line value will override the @file value. See Example 3 below.

It is further specified that if one or more datasets (i.e. positional arguments) are passed on the command line, all files appearing as positional arguments in the “at-file” will be replaced by the one(s) on the command line. See Example 4 below.

In all cases, remember to use the -d option to verify the parsing if you are not sure.

4.4.3.1. Examples

The @file used in the examples, “reducepar”, contains:

# input data files
S20161025S0200.fits
S20161025S0201.fits
S20161025S0202.fits
S20161025S0203.fits
S20161025S0204.fits

# primitive parameters optimization
--param

    # stackFrames
    stackFrames:operation=median

# recipe
-r
    myrecipelib.myrecipe

Example 1: Accumulate a new parameter:

reduce @reducepar --param stackFrames:hsigma=5.0

Summary of parsed options:
--------------------------
Input files: no changes
Parameters: ['stackFrames:operation=median', 'stackFrames:hsigma=5.0']
Recipe: no changes

Example 2: Override a parameter defined in the @file:

reduce @reducepar --param stackFrames:operation=wtmean

Summary of parsed options:
--------------------------
Input files: no changes
Parameters: ['stackFrames:operation=wtmean']
Recipe: no changes

Example 3: Override the recipe:

reduce @reducepar -r myrecipelib.different_recipe

Summary of parsed options:
--------------------------
Input files: no changes
Parameters: no changes
Recipe: myrecipelib.different_recipe

Example 4: Override the input files. All the files in the @files will be ignored:

reduce @reducepar S20161025S0111.fits

Summary of parsed options:
--------------------------
Input files: S20161025S0111.fits
Parameters: no changes
Recipe: no changes