pytest2md 20200601

Create and run markdown Readmes from within pytest

---

author: gk version: 20200601

---

Generating Markdown - While Testing Contained Claims

Table Of Contents

• Inline Python Function Execution
• Features
  • html_table
  • sh_file
  • bash_run
  • md_from_source_code
  • Table of Contents
• Link Replacements
  • Spec
  • Code Repo Hoster Specific Source Links
    • Setting a link template
    • Link Refs
  • Summary
• Tips
• Isolation

Few things are more annoying than stuff which does not work as announced, especially when you find out only after an invest of time and energy.

Documentation is often prone to produce such situations, since hard to keep 100% in sync with the code evolution.

This is a set of tools, generating documentation, while verifying the documented claims about code behaviour - without the need to adapt the source code, e.g. by modifying doc strings:

When the documentation is using a lot of code examples then a very welcome additional benefit of writing it like shown is the availability of source code autoformatters.

Other Example:

This "README.md" was built into this template, where html comment style placeholders had been replaced while running pytest on test_tutorial.

Lets run a bash command and assert on its results. Note that the command is shown in the readme, incl. result and the result can be asserted upon.

$ cat "/etc/hosts" | grep localhost
127.0.0.1   axc3.axiros.com localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6

$ ls "/home/gk/repos/pytest2md/tests"
assets
__pycache__
test_basics.py
test_changelog.py
test_tutorial.py

$ ls -lta /etc/hosts
-rw-r--r--. 1 root root 308 May  8 23:47 /etc/hosts

---

Generated by:

Inline Python Function Execution

via the md_from_source_code function you can write fluent markdown (tested) python combos:

hi = 'hello world'
assert 'world' in hi
print(hi.replace('world', 'joe'))

The functions are evaluated and documented in the order they show up within the textblocks.

Please keep tripple apostrophes - we split the text blocks, searching for those.

State is kept within the outer pytest function, like normally in python. I.e. if you require new state, then start a new pytest function.

Stdout is redirected to an output collector function, i.e. if you print this does result in an "Output" block. If the printout starts with "MARKDOWN:" then we won't wrap that output into fenced code blocks but display as is.

If the string 'breakpoint' occurs in a function body, we won't redirect standardout for displaying output.

Features

html_table

ht = p2m.html_table

print(ht([['foo', 'bar'], ['bar', 'baz']], ['h1', 'h2']))
print('As details when summary arg is given:')
t = ht(
    [['joe', 'doe'], ['suzie', 'wong']],
    ['first', 'last'],
    summary='names. click to open...',
)
assert 'details' in t
assert 'joe</td' in t
print(t)

sh_file

$ cat "test_file.json"
{
    "a": [
        {
            "testfile": "created"
        },
        "at",
        "Wed Jun  3 19:37:58 2020"
    ]
}

as details

$ cat "test_file.json"
{
    "a": [
        {
            "testfile": "created"
        },
        "at",
        "Wed Jun  3 19:37:58 2020"
    ]
}

test_file.json

as link

---

Generated by:

# if content is given it will create it:
p2m.sh_file(fn, lang='javascript', content=c)
# summary arg creates a details structure:
p2m.sh_file(fn, lang='javascript', content=c, summary='as details')
# creates a link (say True to have the filename as link text)
p2m.sh_file(fn, lang='javascript', content=c, as_link=True)
p2m.sh_file(fn, lang='javascript', content=c, as_link='as link')

bash_run

$ ./some_non_existing_command_in_assets arg1
/bin/sh: /home/gk/repos/pytest2md/tests/assets/some_non_existing_command_in_assets: No such file or directory

$ ls -lta | grep total | head -n 1
total 84

$ ls -lta
total 84
drwxr-xr-x.  9 gk gk  4096 Jun  3 19:37 .
-rw-r--r--.  1 gk gk  2198 Jun  3 19:37 .README.tmpl.md
drwxr-xr-x.  3 gk gk  4096 Jun  3 19:35 pytest2md
drwxrwxr-x.  2 gk gk  4096 Jun  3 15:20 pytest2md.egg-info

...(output truncated - see link below)

• Output of ls -lta

$ ls -lta

ls -lta

---

Generated by:

run = partial(p2m.bash_run, cmd_path_from_env=True)
# by default we search in normal environ for the command to run
# but we provide a switch to search in test assets.
# errors are redirected to stdout
res = p2m.bash_run(
    'some_non_existing_command_in_assets arg1',
    cmd_path_from_env=False,
    ign_err=True,
)
assert res[0]['exitcode'] != 0
res = run('ls -lta | grep total | head -n 1')
assert 'total' in res[0]['res']
res = run('ls -lta', into_file='bash_run.txt')
assert 'total' in res[0]['res']
# Ending with .html it converts ansi escape colors to html:
# simple link is created.
# (requires pip install ansi2html)
run('ls -lta', into_file='bash_run.html')

md_from_source_code

Inserted markdown from running python.

Strings in double apos. are rendered, no need to call a render function.

md('Inserted markdown from running python.')
print('From output of running python ')

Output:

From output of running python

More markdown

print('From another function')

Output:

From another function

Strings can also contain instructions, like this (looked up in p2m.MdInline namespace class)

$ cd /etc; ls -lta | head -n 5
total 2716
drwxr-xr-x. 165 root root     12288 Jun  3 09:04 .
-rw-r--r--.   1 root root    139649 Jun  3 09:04 ld.so.cache
-rw-r--r--.   1 root root        67 Jun  3 07:44 resolv.conf
drwxr-xr-x.   2 root root      4096 Jun  1 19:02 alternatives

Default inline functions (add your own in module headers):

print(
    [
        k
        for k in dir(pytest2md.MdInline)
        if not k.startswith('_')
    ]
)

Output:

['bash', 'sh_file']

---

Generated by

        def some_test_function():
            """
            Strings in double apos. are rendered, no need to call a render function.
            """

            def func1():
                md('Inserted markdown from running python.')
                print('From output of running python ')

            """
            > More  markdown
            """

            def func2():
                print('From another function')

            """
            Strings can also contain instructions, like this (looked up in p2m.MdInline namespace class)

            <bash: cd MY_REPL_DIR; ls -lta | head -n 5>

            Default inline functions (add your own in module headers):
            """

            def known():
                print(
                    [
                        k
                        for k in dir(pytest2md.MdInline)
                        if not k.startswith('_')
                    ]
                )

            # repl dict simply replaces keys with values before any processing:
            p2m.md_from_source_code(repl={'MY_REPL_DIR': '/etc'})

Table of Contents

p2m.write_markdown(with_source_ref=True, make_toc=True)

See this tutorial.

Link Replacements

Technical markdown content wants to link to source code often. How to get those links working and that convenient?

The module does offer also some source finding / link replacement feature, via the mdtool module. The latter link was built simply by this:

[mdtool]<SRC>

Other example: This test_tutorial.py link was created by replacing "SRC" with the path to a file matching, under a given directory, prefixed by an arbitrary base URL.

Spec

These will be replaced:

[title:this,fmatch:test_tutorial,lmatch:line_match] <SRC> (remove space between] and <)

• title: The link title - text the user reads
• fmatch: substring match for the link destination file
• lmatch: Find matching line within that file
• show_raw: Link to raw version of file, not the one rendered by the repo server
• path: Fix file path (usually derived by fmach)
• line: Fix the line number of the link (usually done via lmatch)

Code Repo Hoster Specific Source Links

Github, Gitlab, Bitbucked or Plain directory based static content servers all have their conventional URLs regarding those links.

Since all of these are just serving static content w/o js possibilities, you have to parametrize the intended hoster in your environment, before running a pytest / push cycle. That way the links will be working on the hoster.

Currently we understand the following namespaces for links:

{
    "github": "https://github.com/%(gh_repo_name)s/blob/%(git_rev)s/%(path)s%(line:#L%s)s",
    "github_raw": "https://raw.githubusercontent.com/%(gh_repo_name)s/%(git_rev)s/%(path)s%(line:#L%s)s",
    "static": "file://%(d_repo_base)s/%(path)s",
    "static_raw": "file://%(d_repo_base)s/%(path)s"
}

Setting a link template

• export MD_LINKS_FOR=github # before running pytest / push
• <!-- md_links_for: github --> # in the markdown template, static

The latter can be overwritten by environ, should you want to push from time to time to a different code hoster.

Link Refs

We minimize the problem of varying generated target markdown, dependent on the hoster. How? Like any problem in IT is solved.

By building reference links the differences of e.g. a README.md for github vs. gitlab is restricted to the links section on the end of the generated markdown. In the markdown bodies you'll just see link names, which remain the same.

Check the end of the rendering result at the end of this README.md, in order to see the results for the hoster you are reading this markdown file currently.

Summary

• At normal runs of pytest, the link base URL is just a local file:// link,

• Before pushes one can set via environ (e.g. export MD_LINKS_FOR=github) these e.g. to the github base URL or the repo.

• [key-values] constructs are supported as well, extending to beyond just the base url. Example following:

Source code showing is done like this:

    def test_sh_code(self):
        md('Source code showing is done like this:')
        p2m.sh_code(self.test_sh_code)
        md(
            '> Is [title:this,fmatch:test_tutorial,lmatch:exotic]<SRC> an exotic form of a recursion? ;-)  '
        )

Is this an exotic form of a recursion? ;-)

Command Summary

cat "/etc/hosts" | grep localhost
ls "/home/gk/repos/pytest2md/tests"
ls -lta /etc/hosts
/home/gk/repos/pytest2md/tests/assets/some_non_existing_command_in_assets arg1
ls -lta | grep total | head -n 1
ls -lta
ls -lta

*Auto generated by pytest2md, running ./tests/test_tutorial.py

Tips

• Skip "inner" functions, except matching ones: export P2RUN=my_func_name_match

• Turn off stdout redirection: export P2MFG=true (for breakpoints in tested modules)

• Local Renderer:

  pip install grip

to get a local github compliant markdown renderer, reloading after changes of the generated markdown.

• Using fixtures with unittest style test classes

Create a conftest.py like:

root@localhost tests]# cat conftest.py
import pytest
import pytest2md as p2m


@pytest.fixture(scope='class')
def write_md(request):
    def fin():
        request.cls.p2m.write_markdown(with_source_ref=True, make_toc=False)

    request.addfinalizer(fin)

then use like:

@pytest.mark.usefixtures('write_md')
class TestDevUsage(unittest.TestCase):
    p2m = p2m # on module level: p2m.P2M(__file__, fn_target_md='README.md')

    (...)

Isolation

None. If you would screw up your host running pytest normally, then you will get the same result, when running markdown generating tests.

---

Here is a bigger tutorial, from pytest2md.