Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-5790

Provide unified plotting abstraction layer

    XMLWordPrintable

    Details

    • Type: Epic
    • Status: To Do
    • Resolution: Unresolved
    • Fix Version/s: None
    • Component/s: afw
    • Epic Name:
      plotting abstraction layer

      Description

      The stack has dozens of "import matplotlib" statements scattered around. This can cause unexpected behavior, slowdowns and can mess up subsequent matplot commands (e.g. if a implicit import prevents a subsequent explicit import from setting the backend).

      We provide an afw.display abstraction for doing ds9-related things, so we should also have an afw.plot or similar abstraction for plotting. One could then swap in Seaborn or another plotting system more easily, and it would reduce the "import matplotlib" to one place, which would only occur when e.g. that class is instantiated.

      A first step of this would be a simple class that just handles the import and environment setup, while not actually abstracting away any of the matplotlib calls. We could then make it a more general abstraction layer if desired.

        Attachments

          Issue Links

          is triggered by

          Bug - A problem which impairs or prevents the functions of the product. DM-3889 Multi-threaded matplotlib test issue in meas_algorithms

          • Done

          Story - Created by Jira Software - do not edit or delete. Issue type for a user story. DM-4752 Build on Mac very slow due to running fc-list

          • Done
          relates to

          Bug - A problem which impairs or prevents the functions of the product. DM-33437 Eups distrib source installs of Spectractor (in Linux, in a desktop env) create interactive matplotlib windows

          • Done

          Bug - A problem which impairs or prevents the functions of the product. DM-8656 meas_astrom tests depend on PyQt4 and Qt4 and break with PyQt5 and Qt5

          • Done

          Bug - A problem which impairs or prevents the functions of the product. DM-14159 Wrap matplotlib use in meas_mosaic

          • Done

          Story - Created by Jira Software - do not edit or delete. Issue type for a user story. DM-7444 Improve jointcal plotting backend

          • Done

          Story - Created by Jira Software - do not edit or delete. Issue type for a user story. DM-8676 Hide matplotlib imports in meas_algorithms

          • Done

          Bug - A problem which impairs or prevents the functions of the product. DM-2588 Catch RuntimeError on matplotlib.pyplot import and remove more unprotected/unnecesary pyplot imports

          • Done

          Improvement - An improvement or enhancement to an existing feature or task. DM-4096 Please unify the various scripts for displaying cameras

          • Won't Fix
          mentioned in

          Page Loading...

            Activity

            Ascending order - Click to sort in descending order
            Hide
            Permalink
            price Paul Price added a comment -

            This is a worthy goal, but given all the different kinds of plots there are I think it would be difficult to do well. I strongly suggest you RFC this before starting anything to minimise frustration.

            Show
            price Paul Price added a comment - This is a worthy goal, but given all the different kinds of plots there are I think it would be difficult to do well. I strongly suggest you RFC this before starting anything to minimise frustration.
            Hide
            Permalink
            Parejkoj John Parejko added a comment -

            An RFC with an example interface is a good idea.

            I think if nothing else, we want a single place to import and setup matplotlib from (which could come with an LSST "stamp of approval" set of font sizes, etc.). Even if it's not a fully-abstract system, something that lets us safely do something like this would be a big help:

            if self.plot:
            		 
                plt = afw.Matplotlib()
            		 
                plt.scatter(matplotlib_stuff)

            Show
            Parejkoj John Parejko added a comment - An RFC with an example interface is a good idea. I think if nothing else, we want a single place to import and setup matplotlib from (which could come with an LSST "stamp of approval" set of font sizes, etc.). Even if it's not a fully-abstract system, something that lets us safely do something like this would be a big help: if self.plot: plt = afw.Matplotlib() plt.scatter(matplotlib_stuff)
            Hide
            Permalink
            jsick Jonathan Sick added a comment -

            Also keep in mind this prior art regarding a unified plotting abstraction layer:

            https://vega.github.io and https://vega.github.io/vega-lite/ (JSON Schema / visualization grammar).

            Show
            jsick Jonathan Sick added a comment - Also keep in mind this prior art regarding a unified plotting abstraction layer: https://vega.github.io and https://vega.github.io/vega-lite/ (JSON Schema / visualization grammar).
            Hide
            Permalink
            Parejkoj John Parejko added a comment - - edited

            Related to this (since I'm having trouble setting the backend in my own code), I just tried some quick matplotlib hacking we'll have to to take care of any exposed matplotlib.use(), import matplotlib.pyplot, import matplotlib.pylab, or import matplotlib.backends calls as well. Some of that work was done as part of DM-8656, but there may be more such imports lurking.

            Show
            Parejkoj John Parejko added a comment - - edited Related to this (since I'm having trouble setting the backend in my own code), I just tried some quick matplotlib hacking we'll have to to take care of any exposed matplotlib.use() , import matplotlib.pyplot , import matplotlib.pylab , or import matplotlib.backends calls as well. Some of that work was done as part of DM-8656 , but there may be more such imports lurking.
            Hide
            Permalink
            tjenness Tim Jenness added a comment -

            What do we think we should do with this ticket?

            Show
            tjenness Tim Jenness added a comment - What do we think we should do with this ticket?
            Hide
            Permalink
            Parejkoj John Parejko added a comment -

            I do still wish we had this, but I suspect this is not at all a priority (although DM-33437 is a reminder that matplotlib imports can be a problem in many ways).

            Show
            Parejkoj John Parejko added a comment - I do still wish we had this, but I suspect this is not at all a priority (although DM-33437 is a reminder that matplotlib imports can be a problem in many ways).

              People

              Assignee:
              Unassigned Unassigned
              Reporter:
              Parejkoj John Parejko
              Watchers:
              David Shupe, John Parejko, John Swinbank, Jonathan Sick, Michael Wood-Vasey, Paul Price, Simon Krughoff, Tim Jenness
              Votes:
              0 Vote for this issue
              Watchers:
              8 Start watching this issue

                Dates

                Created:
                Updated:

                  Jenkins

                  No builds found.