John Snow <jsnow@redhat.com> writes:
> Change get_doc_indented() to preserve indentation on all subsequent text
> lines, and create a compatibility dedent() function for qapidoc.py to
> remove that indentation. This is being done for the benefit of a new
Suggest "remove indentation the same way get_doc_indented() did."
Aight.
> qapidoc generator which requires that indentation in argument and
> features sections are preserved.
>
> Prior to this patch, a section like this:
>
> ```
> @name: lorem ipsum
> dolor sit amet
> consectetur adipiscing elit
> ```
>
> would have its body text be parsed as:
Suggest "parsed into".
Why? (I mean, I'll do it, but I don't see the semantic difference personally)
> (first and final newline only for presentation)
>
> ```
> lorem ipsum
> dolor sit amet
> consectetur adipiscing elit
> ```
>
> We want to preserve the indentation for even the first body line so that
> the entire block can be parsed directly as rST. This patch would now
> parse that segment as:
If you change "parsed as" to "parsed into" above, then do it here, too.
>
> ```
> lorem ipsum
> dolor sit amet
> consectetur adipiscing elit
> ```
>
> This is helpful for formatting arguments and features as field lists in
> rST, where the new generator will format this information as:
>
> ```
> :arg type name: lorem ipsum
> dolor sit amet
> consectetur apidiscing elit
> ```
>
> ...and can be formed by the simple concatenation of the field list
> construct and the body text. The indents help preserve the continuation
> of a block-level element, and further allow the use of additional rST
> block-level constructs such as code blocks, lists, and other such
> markup. Avoiding reflowing the text conditionally also helps preserve
> source line context for better rST error reporting from sphinx through
> generated source, too.
What do you mean by "reflowing"?
Poorly phrased, was thinking about emacs too much. I mean munging the text post-hoc for the doc generator such that newlines are added or removed in the process of re-formatting text to get the proper indentation for the new rST form.
In prototyping, this got messy very quickly and was difficult to correlate source line numbers across the transformation.
It was easier to just not munge the text at all instead of munging it and then un-munging it.
(semantic satiation: munge munge munge munge.)
> This understandably breaks the existing qapidoc.py; so a new function is
> added there to dedent the text for compatibility. Once the new generator
> is merged, this function will not be needed any longer and can be
> dropped.
>
> I verified this patch changes absolutely nothing by comparing the
> md5sums of the QMP ref html pages both before and after the change, so
> it's certified inert. QAPI test output has been updated to reflect the
> new strategy of preserving indents for rST.
I think the remainder is unnecessary detail. Drop?
As long as you're convinced it's safe, it's done its job and we thank it for its service
🫡
> before:
>
> 69cde3d6f18b0f324badbb447d4381ce manual_before/interop/qemu-ga-ref.html
> 446e9381833def2adc779f1b90f2215f manual_before/interop/qemu-qmp-ref.html
> df0ad6c26cb4c28b85d663fe44609c12 manual_before/interop/qemu-storage-daemon-qmp-ref.html
>
> after:
>
> 69cde3d6f18b0f324badbb447d4381ce manual/interop/qemu-ga-ref.html
> 446e9381833def2adc779f1b90f2215f manual/interop/qemu-qmp-ref.html
> df0ad6c26cb4c28b85d663fe44609c12 manual/interop/qemu-storage-daemon-qmp-ref.html
>
> Signed-off-by: John Snow <jsnow@redhat.com>
> ---
> docs/sphinx/qapidoc.py | 29 ++++++++++++++++++++++++-----
> scripts/qapi/parser.py | 5 +++--
> tests/qapi-schema/doc-good.out | 32 ++++++++++++++++----------------
> 3 files changed, 43 insertions(+), 23 deletions(-)
>
> diff --git a/docs/sphinx/qapidoc.py b/docs/sphinx/qapidoc.py
> index e675966defa..f2f2005dd5f 100644
> --- a/docs/sphinx/qapidoc.py
> +++ b/docs/sphinx/qapidoc.py
> @@ -26,6 +26,7 @@
>
> import os
> import re
> +import textwrap
>
> from docutils import nodes
> from docutils.parsers.rst import Directive, directives
> @@ -53,6 +54,21 @@
> __version__ = "1.0"
>
>
> +def dedent(text: str) -> str:
> + # Temporary: In service of the new QAPI Sphinx domain, the QAPI doc
> + # parser now preserves indents in args/members/features text.
> + # QAPIDoc does not handle this well, so undo that change here.
A comment should explain how things are. This one explains how things
have changed. Suggest:
# Adjust indentation to make description text parse as paragraph.
If we planned to keep this, we might want to explain in more detail, as
I did in review of v1. But we don't.
> +
> + lines = text.splitlines(True)
> + if re.match(r"\s+", lines[0]):
> + # First line is indented; description started on the line after
> + # the name. dedent the whole block.
> + return textwrap.dedent(text)
> +
> + # Descr started on same line. Dedent line 2+.
> + return lines[0] + textwrap.dedent("".join(lines[1:]))
> +
> +
> # Disable black auto-formatter until re-enabled:
> # fmt: off
>
> @@ -176,7 +192,7 @@ def _nodes_for_members(self, doc, what, base=None, branches=None):
> term = self._nodes_for_one_member(section.member)
> # TODO drop fallbacks when undocumented members are outlawed
> if section.text:
> - defn = section.text
> + defn = dedent(section.text)
> else:
> defn = [nodes.Text('Not documented')]
>
> @@ -214,7 +230,7 @@ def _nodes_for_enum_values(self, doc):
> termtext.extend(self._nodes_for_ifcond(section.member.ifcond))
> # TODO drop fallbacks when undocumented members are outlawed
> if section.text:
> - defn = section.text
> + defn = dedent(section.text)
> else:
> defn = [nodes.Text('Not documented')]
>
> @@ -249,7 +265,7 @@ def _nodes_for_features(self, doc):
> dlnode = nodes.definition_list()
> for section in doc.features.values():
> dlnode += self._make_dlitem(
> - [nodes.literal('', section.member.name)], section.text)
> + [nodes.literal('', section.member.name)], dedent(section.text))
> seen_item = True
>
> if not seen_item:
> @@ -272,9 +288,12 @@ def _nodes_for_sections(self, doc):
> continue
> snode = self._make_section(section.tag)
> if section.tag and section.tag.startswith('Example'):
> - snode += self._nodes_for_example(section.text)
> + snode += self._nodes_for_example(dedent(section.text))
> else:
> - self._parse_text_into_node(section.text, snode)
> + self._parse_text_into_node(
> + dedent(section.text) if section.tag else section.text,
> + snode,
> + )
> nodelist.append(snode)
> return nodelist
>
> diff --git a/scripts/qapi/parser.py b/scripts/qapi/parser.py
> index 7b13a583ac1..43167ef0ab3 100644
> --- a/scripts/qapi/parser.py
> +++ b/scripts/qapi/parser.py
> @@ -437,6 +437,7 @@ def _match_at_name_colon(string: str) -> Optional[Match[str]]:
> return re.match(r'@([^:]*): *', string)
>
> def get_doc_indented(self, doc: 'QAPIDoc') -> Optional[str]:
> + """get_doc_indented preserves indentation for later rST parsing."""
A proper function comment explains what the function does. This one
merely points out one minor aspect. Easy fix: delete it. Alternative
fix: write a proper function comment.
Fair. I'll take a stab at writing a better one, but encourage inevitable tweaks to the phrasing to be handled by the maintainer afterwards ;)
> self.accept(False)
> line = self.get_doc_line()
> while line == '':
[...]
Just commit message and doc nitpicks, so
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Thanks! I'll make some good faith adjustments and keep the RB, but re-review before you merge and (please) feel free to make phrasing tweaks as you see fit.