20 years with Gettext

Experiences from the PostgreSQL project

Peter Eisentraut

Who am I?

• FOSS claim to fame: PostgreSQL
• C programmer
• translation toolsmith
• translator

What is PostgreSQL?

• relational (SQL) database management system
  • 1984: started
  • 1996: open source
• ca. 1 million lines of code (C, Perl, make, autoconf)
• yearly major releases (12, 13, 14, 15, ...)
  • 5 years community maintenace

Translation in PostgreSQL

• standard (GNU?) Gettext
• currently about 17 languages
• 26 message catalogs
• ca. 8000 strings
• 5–6 branches
• separate Git repo
• some related projects have their own workflows (pgAdmin, JDBC driver)
• documentation translation is handled separately

https://babel.postgresql.org/

current workflow

• developer marks up messages: _("foo")
  • mostly automatic: e.g., log_error("foo")
• web site runs xgettext, msgmerge
• translator downloads, translates locally (any tool)
• translator commits or sends in
• all translations copied to source code repo at release time

current toolchain

• GNU Gettext
• Autoconf/GNU Make
  • coming soon: Meson
• local editor of choice
  • some teams use Crowdin
• bag of shell and Perl scripts

translation pros

• implicit review system of all program messages
• also a recruiting system for code contributors

translation challenges: content

• Challenges to recuit translators/translations:
  • PostgreSQL is used by technically minded users
  • User community is smaller
  • Terminology is specialized
  • Workflow is pretty obscure

translation challenges: strings

"expected %d things but found %d things"

"cannot apply a %s with a %s"

translation challenges: build system

• keep forgetting to add new files in the right places
• handling shared files is complicated

translation challenges: Gettext APIs

• locale settings
• encoding conversion
• multiple languages per process
• multiple languages per output

translation challenges: Gettext tools

• slowish
• pre-SCM
• backpatching cumbersome

Gettext toolchain maintenance

The future for PostgreSQL

• current focus is on Meson transition
• also ICU adoption
• need to modernize ... Weblate?, OmegaT?
• keep recruiting and retaining contributors

The future for Gettext?

• Is there something different we should be looking at?

Conclusions & Links