pystitch / stitch
- пятница, 2 сентября 2016 г. в 03:14:04
Python
Write reproducible reports in Markdown
A knitr- RMarkdown-like library, in Python.
Note: You might want to consider Jan Schulz's knitpy instead. It's probably more mature at this point. However, I wanted to see if there was a simpler way of doing things.
The high-level goal of this type of library (knitr/RMarkdown, knitpy, and stitch) is to make writing reproducible reports easier.
Documentation is available on readthedocs.
See the project's website for a side-by-side comparison of input markdown and stitched HTML.
More complex examples are linked to from there as well.
At the moment stitch
can be installed from pip via
pip install knotr
I know, it's confusing.
I've filed a claim for stitch
on PyPI, but I think the people working that support queue are over-worked.
Once that gets processed, I'll put it up on conda-forge as well.
If you need a mnemonic, it's "I want knitr, but not the one written in R."
Also I wanted to confuse R users.
And knots are kind of like a buggy version of knits.
The goal was to keep stitch
itself extremely simple by reusing
existing libraries. A high level overview of our tasks is
Fortunately the building blocks are all there.
We reuse
So all stitch
has to do is to provide a command-line interface, scan
the document for code chunks, manage some kernels, hand the code to the
kernels, pass the output to an appropriate pandocfilter
.
The biggest departure from knitpy
is the use of pandoc's JSON AST.
This is what you get from pandoc -t json input.md
This saves us from having do any kind of custom parsing of the markdown.
The only drawback so far is somewhat inscrutable Haskell exceptions if
stitch
happens to produce a bad document.