Big picture workflow

[hadley]

(@MichaelChirico these are the notes I made from our conversation, updated based on my experience with the PRs)

A few scenarios to consider:

• Big package with large community (e.g. data.table, tidyverse): few developers, many translators.

• Bilingual author who wants to provide messages in English + their native language (e.g. metR): developer = translator

• One really passionate user or a language community who contributes to many packages.

Basic workflow

Initial translation:

• po_scan(): R/C source -> .pot

• po_create(): .pot -> .po

• Translators edit .po

• po_compile(): .po -> .mo

Updates:

• po_scan() to update .pot

• po_update() to update .po

• Translators update .po

• po_compile(): .po -> .mo

More details

Getting started

• Add tr_() and tr_n() functions.

  • Should also have untr_() or similar to declare that you have reviewed a message and have decided not to translate it?

• Do whatever is needed on the C side

• Run po_report() and pick a function to translate. Repeat.

  • This would look at all common messaging functions (cat()/message()/warning()/stop() and rlang equivalents) and check that the first argument is a call to tr_() or tr_n().

• Use po_scan() to scan R/ and src/ to generate .pot files

Translator: Adding a new language

• Run po_create("fr")

  • Should we track GitHub user name of contributors? Would be nice to be able to ping folks when translations need updating.

• Fill in translations

• Run po_compile() to build .mo file

  • Probably best done with CI --- probably in the pull request or on merge to main. Or maybe because it's a binary, maybe do it on release? (e.g. is ~ 1 meg for data.table). If on release, what's the workflow for prototyping.

  • Would be nice to be able to dynamically get just the translations you need.

Developer: Translating more messages

• Run po_report() and pick an function to translate. Repeat.

• Use po_update() to update individual .po files.

• Ask translators to update.

Before release

• Use po_report() to make sure all messages have translations, and all languages have translations for message.

  • use_release_issue() should add to bullets if po directory exists.

• As needed, ping translators to request updates

• po_compile() to make sure all up to date

[MichaelChirico]

Should we track GitHub user name of contributors? Would be nice to be able to ping folks when translations need updating.

From what I understand, this would be served well by X- fields in the .po metadata:

https://github.com/wch/r-source/blob/6bb3d143d8a4e8a052c9a8c180f5d419beeee1db/src/library/base/po/R-it.po#L19

e.g. we could add X-GitHub-Id: michaelchirico and read from that in whatever tooling we add