Use Jinja i18n to gettext instead of databags

[johnzhou721]

What is the problem or limitation you are having?

Use Jinja i18n to localize the recurring phrases, not data bags, this provides more flexibility

Describe the solution you'd like

Jinja i18n

Describe alternatives you've considered

Status quo and put the {{}} in the databags but people can inject arbitrary code into translations.

Additional context

Not sure how to use Jinja plugins in lektor

[johnzhou721]

Hmm... Lektor i18n plugin actually supports doing the template files. And guess what it uses.... [drumroll] Jinja2.ext.i18n!!!

@freakboy3742 Would you prefer the template messages to be separate though? Also would you mind if I remove the keys and just use english strings as msgids?

[freakboy3742]

If I'm understanding you correctly - yes: English keys would be preferable. The databags are using msgids purely because that's what was suggested in the docs (at the time, anyway... don't know if that's still true).

[johnzhou721]

@freakboy3742 Sorry, another question: Would you prefer the po files for the template messages to be separate from that of the main text, or is all-in-one po file okay w/ you?

[freakboy3742]

One PO file would be preferable, but we need to make sure that whatever is putting new keys into the PO file will kept them in sorted location order, and that location order needs to merge nicely with what the rest of the website keys generate. Otherwise we're going to be in a world of pain whenever we add/remove keys.

If using 2 PO files is the only way to make that happen without major effort, then it probably not worth that effort.

[johnzhou721]

@freakboy3742 Got it.

[johnzhou721]

@freakboy3742 Progress: beeware/lektor-i18n-plugin#6 guarantees this.

[johnzhou721]

In case I didn't make it clear: I am working on this.

[johnzhou721]

So: I've mostly got this working, but now I need to sync the databags into the new po files. I've already done this for chinese -- any tips to make the process faster?