pygettext: Extend support for specifying custom keywords

[tomasr8]

Feature or enhancement

Proposal:

pygettext already allows specifying custom keywords in a limited fashion. For example, specifying --keyword=foo will
look for functions named foo in addition to the default keywords (gettext, etc..).

However, the CLI currently only allows specifying single-argument functions (i.e. gettext('foo')). It is not possible to add
keywords that take msgid_plural or msgctxt or both.

Both xgettext and babel support this with a simple keywordspec syntax that defines the function name and argument positions of msgid, msgid_plural and msgctxt.

For example, --keyword:foo:1c,2 defines a pgettext-like function where the msgctxt is the first argument and msgid is the second argument, e.g. foo('context', 'message').

It is also possible to use t to only match functions with that exact number of arguments. For example, foo:1,1t only matches foo('bar') but not foo('bar', 'baz').

I propose to support this in pygettext as well. For that we will need to:

• Support the keyword spec syntax when specifying keywords. I suggest skipping the t specifier for now.
• Support specifying multiple --keyword arguments with the same function name (this will require some internal changes, as those are stored in a dictionary with the function name as the key).
• Support the t specifier.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

Linked PRs

• gh-130453: pygettext: Extend support for specifying custom keywords #130463
• gh-130453: pygettext: Allow overriding default keywords when using --keyword #130709
• gh-130453: pygettext: Allow specifying multiple keywords with the same function name #131380

[tomasr8]

The first PR (with the most changes) was merged, but there is more work needed. I'll reuse this issue for the followup PRs :)

[serhiy-storchaka]

Note also that --keyword without argument and --keyword= resets keywords to default. I am not sure whether we should support this or not.

More important, currently in pygettext the defaults take precedence over --keyword arguments, so you cannot override defaults. This does not look right. How is with this in xgettext?

[tomasr8]

Note also that --keyword without argument and --keyword= resets keywords to default. I am not sure whether we should support this or not.

Not sure either to be honest.

More important, currently in pygettext the defaults take precedence over --keyword arguments, so you cannot override defaults. This does not look right. How is with this in xgettext ?

Both babel and xgettext let you do that. For example for xgettext --keyword='_:1,2' and this input file

_('foo', 'foos', 1)

xgettext will extract

msgid "foo"
msgid_plural "foos"
msgstr[0] ""
msgstr[1] ""

while pygettext will extract

msgid "foo"
msgstr ""

i.e. the default keyword takes precedence. I'll fix it.