Add markdown support.

[daavoo]

Cutting quite a few corners to cover the minimum for iterative/dvclive#91

Proper refactor would make sense as follow-up because the whole package was focused on a single output format (HTML).

---

table: Add generate_markdown.

image: Add generate_markdown.

vega: Add generate_markdown.

"Translate" vega plot to matplotlib figure to save as png.

Introduce render_markdown.

Closes #62

[daavoo]

Leftover from #61

[codecov-commenter]

Codecov Report

Merging #69 (8bce146) into main (70e2da2) will decrease coverage by 1.48%.
The diff coverage is 89.54%.

@@            Coverage Diff             @@
##             main      #69      +/-   ##
==========================================
- Coverage   97.14%   95.65%   -1.49%     
==========================================
  Files          16       19       +3     
  Lines         491      622     +131     
  Branches       77       90      +13     
==========================================
+ Hits          477      595     +118     
- Misses          8       22      +14     
+ Partials        6        5       -1     

Impacted Files	Coverage Δ
src/dvc_render/markdown.py	60.60% <60.60%> (ø)
src/dvc_render/table.py	81.81% <90.90%> (+8.48%)	⬆️
src/dvc_render/vega.py	97.18% <92.59%> (-2.82%)	⬇️
src/dvc_render/base.py	100.00% <100.00%> (ø)
src/dvc_render/exceptions.py	100.00% <100.00%> (ø)
src/dvc_render/html.py	87.50% <100.00%> (+3.50%)	⬆️
src/dvc_render/image.py	100.00% <100.00%> (ø)
src/dvc_render/plotly.py	100.00% <100.00%> (ø)
src/dvc_render/utils.py	100.00% <100.00%> (ø)
tests/test_image.py	100.00% <100.00%> (ø)
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 70e2da2...8bce146. Read the comment docs.

[pared]

Maybe we should require users to install altair to support all vega plots in md?

[daavoo]

Maybe we should require users to install altair to support all vega plots in md?

I initially considered it but decided to go with matplotlib mainly because:

• matplotlib covers the DVCLive<>CML use case (in the future we need to revisit this and consider altair for cover all use cases)

• even though altair and matplotlib are similarly heavier, matplotlib is much popular among ML and as such more likely to be already installed in the user env. For altair it feels like we could be the only reason to install a heavy dependency in the env

• In order to save images, altair requires additional dependency altair_saver which in turn requires a even heavier non-python dependency:

  To be used, it requires the Selenium Python package, and a properly configured installation of either chromedriver or geckodriver.

[pared]

I now wonder if this use case is not another point for dvc-render to have some support for javascript. If there is some pythonic wrapper for js maybe we could leverage vega to produce images in non-linear template use cases. But I guess that would require some research to determine this idea's feasibility.