Two Seemingly Untranslated Strings

[johnzhou721]

Describe the bug

See: /[language]/community/behavior/code-of-conduct/, summary not translated

See: https://beeware.org/`[language]`/, Contribute to BeeWare button not translated but is extracted into the PO files.

[EDIT html markup]

Steps to reproduce

Goto specified

Expected behavior

Translat{ed,able} strings.

Screenshots

No response

Environment

Irrelevant

Logs

No response

Additional context

No response

[johnzhou721]

The contribute to BeeWare button is because of line_starts_new_block in lektor_i18n_plugin and it checks if the previous line has a colon or not to know if it's a heading or a start of line.

It gets detected as a Markdown header deliminator, so it's translatable, and so the translatable string becomes something like

Contribute to BeeWare
---
anotherkey: value

rather than just Contribute to BeeWare, causing gettexting to fail.

The problem is not present during extraction because they used lektor's own functions etc to parse a bunch of things and somehow evaded this bug. I think a more long-term project is to use more of lektor's functions to parse the file rather than throwing our own ad-hoc solution when substituting translated strings.

Empirical testing suggests that only if you have 3 -s does the syntax work correctly, not an arbitary amount of dashes as seen in the lektor-i18n-plugin code.

How I found this out, plus more tests to prove this

This heuristic is extremely unreliable. See the following random lr file:

title: Welcome to tangledtest!
---
body:

This is a basic demo website that shows how to use Lektor for a basic
website with some pages and a blog.

Some random notes regarding disney princesses.

Tangled: The Series
--------------------

It is better than Frozen II.
---
randomkey: Some stuff.

The dashes before randomkey: Some stuff actually would probably get extracted as a Markdown heading, while Tangled: The Series is extracted as a block seperation!

Try changing the 3 dashes to any amount of dashes just after the Frozen II line and it'll be a section heading, not a field seperation.

Therefore, I am going to change the heuristic into simply checking 3 dashes to know if a new field begins, as changing the amount of dashes below title: actually make none of the body display... so I think this is enough testing to assume that the number of dashes is what dicriminates between a markdown section and a block/field seperation is the number of dashes.

The second part of this issue is still pending investigation.

[johnzhou721]

No... turns out, in button blocks, this magic value is 4, not 3... back to the drawing board, I guess.

[johnzhou721]

Well... I've implemented a good enough heuristic for the first issue based on https://www.getlektor.com/docs/api/db/types/flow/ now.