podman/hack/markdown-preprocess

#!/usr/bin/env python3
#
# markdown-preprocess - filter *.md.in files, convert to .md
#
"""
Simpleminded include mechanism for podman man pages.
"""

import filecmp
import glob
import os
import re
import sys

class Preprocessor():
    """
    Doesn't really merit a whole OO approach, except we have a lot
    of state variables to pass around, and self is a convenient
    way to do that. Better than globals, anyway.
    """
    def __init__(self):
        self.infile = ''
        self.pod_or_container = ''
        self.used_by = {}

    def process(self, infile:str):
        """
        Main calling point: preprocesses one file
        """
        self.infile = infile
        # Some options are the same between containers and pods; determine
        # which description to use from the name of the source man page.
        self.pod_or_container = 'container'
        if '-pod-' in infile or '-kube-' in infile:
            self.pod_or_container = 'pod'

        # foo.md.in -> foo.md -- but always write to a tmpfile
        outfile = os.path.splitext(infile)[0]
        outfile_tmp = outfile + '.tmp.' + str(os.getpid())

        with open(infile, 'r', encoding='utf-8') as fh_in, open(outfile_tmp, 'w', encoding='utf-8', newline='\n') as fh_out:
            for line in fh_in:
                # '@@option foo' -> include file options/foo.md
                if line.startswith('@@option '):
                    _, optionname = line.strip().split(" ")
                    optionfile = os.path.join("options", optionname + '.md')
                    self.track_optionfile(optionfile)
                    self.insert_file(fh_out, optionfile)
                # '@@include relative-path/must-exist.md'
                elif line.startswith('@@include '):
                    _, path = line.strip().split(" ")
                    self.insert_file(fh_out, path)
                else:
                    fh_out.write(line)

        os.chmod(outfile_tmp, 0o444)
        os.rename(outfile_tmp, outfile)

    def track_optionfile(self, optionfile: str):
        """
        Keep track of which man pages use which option files
        """
        if optionfile not in self.used_by:
            self.used_by[optionfile] = []
        self.used_by[optionfile].append(self.podman_subcommand('full'))

    def rewrite_optionfiles(self):
        """
        Rewrite all option files, such that they include header comments
        cross-referencing all the man pages in which they're used.
        """
        for optionfile in self.used_by:
            tmpfile = optionfile + '.tmp'
            with open(optionfile, 'r', encoding='utf-8') as fh_in, open(tmpfile, 'w', encoding='utf-8', newline='\n') as fh_out:
                fh_out.write("####> This option file is used in:\n")
                used_by = ', '.join(x for x in self.used_by[optionfile])
                fh_out.write(f"####>   podman {used_by}\n")
                fh_out.write("####> If file is edited, make sure the changes\n")
                fh_out.write("####> are applicable to all of those.\n")
                for line in fh_in:
                    if not line.startswith('####>'):
                        fh_out.write(line)
            # Compare files; only rewrite if the new one differs
            if not filecmp.cmp(optionfile, tmpfile):
                os.unlink(optionfile)
                os.rename(tmpfile, optionfile)
            else:
                os.unlink(tmpfile)

    def insert_file(self, fh_out, path: str):
        """
        Reads one option file, writes it out to the given output filehandle
        """
        # Comment intended to help someone viewing the .md file.
        # Leading newline is important because if two lines are
        # consecutive without a break, sphinx (but not go-md2man)
        # treats them as one line and will unwantedly render the
        # comment in its output.
        fh_out.write("\n[//]: # (BEGIN included file " + path + ")\n")
        with open(path, 'r', encoding='utf-8') as fh_included:
            for opt_line in fh_included:
                if opt_line.startswith('####>'):
                    continue
                opt_line = self.replace_type(opt_line)
                opt_line = opt_line.replace('<<subcommand>>', self.podman_subcommand())
                opt_line = opt_line.replace('<<fullsubcommand>>', self.podman_subcommand('full'))
                fh_out.write(opt_line)
            fh_out.write("\n[//]: # (END   included file " + path + ")\n")

    def podman_subcommand(self, full=None) -> str:
        """
        Returns the string form of the podman command, based on man page name;
        e.g., 'foo bar' for podman-foo-bar.1.md.in
        """
        subcommand = self.infile
        # Special case: 'podman-pod-start' becomes just 'start'
        if not full:
            if subcommand.startswith("podman-pod-"):
                subcommand = subcommand[len("podman-pod-"):]
        if subcommand.startswith("podman-"):
            subcommand = subcommand[len("podman-"):]
        if subcommand.endswith(".1.md.in"):
            subcommand = subcommand[:-len(".1.md.in")]
        return subcommand.replace("-", " ")

    def replace_type(self, line: str) -> str:
        """
        Replace instances of '<<pod string|container string>>' with the
        appropriate one based on whether this is a pod-related man page
        or not.
        """
        # Internal helper function: determines the desired half of the <a|b> string
        def replwith(matchobj):
            lhs, rhs = matchobj[0].split('|')
            # Strip off '<<' and '>>'
            lhs = lhs[2:]
            rhs = rhs[:len(rhs)-2]

            # Check both sides for 'pod' followed by (non-"m" or end-of-string).
            # The non-m prevents us from triggering on 'podman', which could
            # conceivably be present in both sides. And we check for 'pod',
            # not 'container', because it's possible to have something like
            # <<container in pod|container>>.
            if re.match('.*pod([^m]|$)', lhs, re.IGNORECASE):
                if re.match('.*pod([^m]|$)', rhs, re.IGNORECASE):
                    raise Exception(f"'{matchobj[0]}' matches 'pod' in both left and right sides")
                # Only left-hand side has "pod"
                if self.pod_or_container == 'pod':
                    return lhs
                return rhs

            # 'pod' not in lhs, must be in rhs
            if not re.match('.*pod([^m]|$)', rhs, re.IGNORECASE):
                raise Exception(f"'{matchobj[0]}' does not match 'pod' in either side")
            if self.pod_or_container == 'pod':
                return rhs
            return lhs

        return re.sub(r'<<[^\|>]*\|[^\|>]*>>', replwith, line)


def main():
    "script entry point"
    script_dir = os.path.abspath(os.path.dirname(__file__))
    man_dir = os.path.join(script_dir,"../docs/source/markdown")

    try:
        os.chdir(man_dir)
    except FileNotFoundError as ex:
        raise Exception("Please invoke me from the base repo dir") from ex

    # No longer possible to invoke us with args: reject any such invocation
    if len(sys.argv) > 1:
        raise Exception("This script accepts no arguments")

    preprocessor = Preprocessor()
    for infile in sorted(glob.glob('*.md.in')):
        preprocessor.process(infile)

    # Now rewrite all option files
    preprocessor.rewrite_optionfiles()

if __name__ == "__main__":
    main()