workshopsoft/ludic
2
0
Fork 0
Code Issues Pull requests Projects Releases 26 Packages Wiki Activity Actions

compiler: text keys' follow-up (Master's rulings) - a Key in a template literal's hole is an error (the parser marks each hole's string(), TPL_HOLE; write trf(k"...", ...)), trn takes only a plural key kn"..." and tr / trf refuse one (with or without en.po), and a program with no lang line (a package test) uses key literals unchecked and silently - the warning stays for a lang line whose en.po is missing; rejected/i18n_keys gains key_hole (2) and key_rules (3), key_missing's plural moves to trn; LANGUAGE.md and the changeset say so; reseeded, bootstrap-cfree fixpoint holds

Browse source 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Orkun ÇAKILKAYA 2026-09-29 23:46:48 +03:00
parent 66b8e51026
commit 9f9ac4a7ce
13 changed files with 72839 additions and 72164 deletions
Download patch file Download diff file Expand all files Collapse all files

9
LANGUAGE.md
View file

@ -836,7 +836,9 @@ results, list elements and registry values like any other type (`null` is none).
its text after a marker byte - byte 1, or 2 for `kn"..."` - so `k"pause.resume"` is the string
`"\x01pause.resume"`: the runtime's `tr` is a cast, and a translator knows a key from English.
`string(key)` gives that marked text, for a runtime that needs it; a program itself makes text with
`tr`. (A program that declares its own type named `Key` keeps it; the builtin is then out of reach.)
`tr`, and a key in a template literal's hole (`` `at {k}` ``) is an error - write `trf(k"...", ...)`.
`trn` takes a plural key and `tr` / `trf` refuse one, with or without a `lang` line. (A program that
declares its own type named `Key` keeps it; the builtin is then out of reach.)
**The data.** A registry's `@Text` field is text by derivation: its key is
`<registry>.<row>.<field>` - the registry's name in snake case (`Items` -> `items`, `GearKinds` ->
@ -876,9 +878,10 @@ each diagnostic at its `file:line:col`:
- one English under several keys (a split) wants a `#.` description on each, for a translator to
tell them apart: a **warning** at the `.po`'s line.
With no `lang` line, or no source `.po`, nothing is checked, and the first key literal says so once.
With no `lang` line nothing is checked and nothing is said (a package test uses key literals freely); a
`lang` line whose source `.po` is missing checks nothing, and the first key literal says so once.
The source `.po` is gettext's: `msgid` (the key), `msgid_plural` and `msgstr[n]`, `#,` flags
(`fuzzy`), `#.` descriptions, `msgctxt` (read, keying nothing), strings over several lines; `#~`
(`fuzzy`), `#.` descriptions, `msgctxt` (read and ignored: the key is the context), strings over several lines; `#~`
entries are obsolete. `ludic schema` carries a `"lang"` object (null without a `lang` line): every
key used with its kind (`code`, `template`, `data`), its sites, its English, whether it is a plural,
and its description; `unused` (the source's keys used nowhere), `undescribed`, `split`, and per other