# Outline - My path (so far) - Architecture of Matplotlib - How to start contributing --- # Biography - Physics and Math undergrad (Cornell) - Physics PhD (University of Chicago) - Post-doc and then staff at NSLS-II at Brookhaven National Laboratory ---- ## Physics (undergrad) - Started undergrad wanting to be a quantum gravitational theorist - Worked as an under graduate assistant in Sol Gruner's lab building detectors - adapting a pixelated x-ray detector for use in Scanning Transmission Electron Microscope - Became hooked on experimental work - Pivoted goal to be an experimentalist, but still wanted to be faculty ---- ## Physics (grad) - Did experimental soft condensed matter - Jammed colloids (flouresence confocal microscopy) - vapor layer below a Leidenfrost drop (high-speed video) - Connecting theme was work that could not have been done "by hand" - ~TB scale raw data - Became hooked on programming! - switched from MATLAB + c++ -> Python + c++ ---- ## A (failed) job search - Learned in grad school I did not want to be faculty (of any sort) - Did a round of industry job interviews (big tech, a startup, and a small company) - did not get a job - Got the advice "learn to program with others" - Accepted a post-doc at NSLS-II ---- ## Pivot to open source - Went into my post-doc and open source with goal to learn how to work on a software team - Learned how to run and maintain a big code base from Matplotlib team - Started to take on leadership role in Matplotlib - Brought lessons from Matplotlib to NSLS-II ---- ## Research Software Engineer - Not my actual job title - Job role I most closely identify with - Emerging role in science - https://us-rse.org (US) - https://society-rse.org (UK)  ---- ## Bluesky https://blueskyproject.io - Experimental orchestration and data acquisition in Python  --- # Architecture of  Three layers 1. Creation 2. Internal Representation 3. Rendering ---- ## Creation How you tell Matplotlib what to plot ### Explicit API (preferred) ```python= fig, ax = plt.subplots() ax.plot(...) ax.imshow(...) ax.set_xlim(...) ``` ### Implicit API ```python= plt.plot(...) plt.imshow(...) plt.xlim(...) ``` implicit API is implemented with explicit API [users/explain/api_interfaces.html](https://matplotlib.org/stable/users/explain/api_interfaces.html) ---- ## Internal Representation Everything you see has an `Artist` behind it  [tutorials/introductory/quick_start.html#parts-of-a-figure](https://matplotlib.org/stable/tutorials/introductory/quick_start.html#parts-of-a-figure) ---- ## Internal Representation - `Figure` is outer most / root `Artist` - `Figure` has as children `Axes` and `SubFigure` - `Axes` are "plotting area" - have 2 or 3 `Axis` - maintain data <-> screen `Transforms` - hold the 'plotting' `Artists` (`Line2D`, `AxesImage`, `PathCollection`, ...) Truth of what is "in the visualization" is state of these objects ---- ## Internal Representation LIVE DEMO ```python= import matplotlib.pyplot as plt import numpy as np th = np.linpsace(0, np.pi, 512) fig, ax = plt.subplots() ln, = ax.plot(th, np.sin(th)) ln.set_color('purple') ln.set_linewidth(5) ln.set_cloor('xkcd:wine red') ln.set_ydata(np.cos(th)) ... ``` ---- ## Rendering - Mostly hidden from the user - Implements a rough subset of SVG low-level vocabulary - How Matplotlib can save {`png`, `jpg`, `svg`, `pdf`, `ps`, ...} - How Matplotlib manages GUI embedding --- # Matplotlib usage tips ---- ## `subplot_mosaic` ```python= fig = plt.figure() axd = fig.subplot_mosaic('ABD;CCD') ```  [gallery/subplots_axes_and_figures/mosaic.html](https://matplotlib.org/stable/gallery/subplots_axes_and_figures/mosaic.html) https://github.com/ksunden/mosaic_magic_with_matplotlib ---- ## `fig.subfigure` ```python= fig = plt.figure() fig_left, fig_right = fig.subfigures(1, 2) ax_l = fig_left.subplots() axd_r = fig_right.subplot_mosaic('AB;CD') ``` ---- ## Write helpers! - For plotting to "feel fluid" it needs to understand your data structures - Matplotlib is (very) general - Write a wrapper that fits your needs exactly ```python= def my_plotter(ax, data, ...): ... def my_plotter2(fig, data, ...) ... ``` [tutorials/introductory/quick_start.html](https://matplotlib.org/stable/tutorials/introductory/quick_start.html#making-a-helper-functions) https://matplotlib.org/mpl-third-party/ https://github.com/tacaswell/2021-03_APS --- # Getting starting in open source ---- ## Everything helps! - reporting a bug - reporting if you can reproduce a bug - fixing typos in docs - clarifying docs - if the docs confused you but you got it do what you needed...how could the docs be improved? - helping another user ---- ## Work on what you use - Pick a project you use day-to-day - you will have a lot of context (what _should_ it do....) - you will care if the project actually improves! - benefit from your own work - many (many) more projects than the big ones need help! ---- ## Collaborate & work in the open - Publish your code! - Far more likely to find collaborators than competitors - Follow good SWE practices even if you are the only user (so far) - your future self is your most common collaborator, be kind to them - Can have broad and persistent impact ---- ## Collaborate & work in the open http://soft-matter.github.io/trackpy/v0.6.1/ - `trackpy` is the merger of my grad work with Dan Allan's - We met because he found my code and proposed we join forces - He had written the particle location code - I had the track linking code - Gained additional contributors over time - Cited 340+ times, used across diverse domains - Dan is now my closest collaborator and co-worker at NSLS-II ---- ## Get to know the community - software is a team sport - building trust and relationships can take time - always be open to learning ---- ## Time commitment - Get out what you put in - Drive by contributions are OK - Will absorb as much time as you want to give it ---- ## Working in public is terrifying It gets easier with practice ---- ## Resources - https://www.pyopensci.org - https://learn.scientific-python.org/development/