The original reason for the catch-all message was users with bad
data such as decimal intervals, but those get automatically coerced
these days. The common case should now be invalid search strings, which
we can show verbatim.
The majority of the error checking can now be delegated to routines
like parse_f32(), parse_negative_i32() and so on, instead of creating
specific error messages for each type.
- use markdown instead of HTML, to make editing and translating easier
- use a shared prefix
- a few very minor wording tweaks
- we don't need to translate undocumented command errors
- share a string for positive number of days
- share a string for invalid property and state arguments, and avoid
listing them out
Related discussion: https://github.com/ankitects/anki/pull/922
https://github.com/ankitects/anki/pull/889
- We can't use compare_locales's getChecker(), as it seems to
be focused on checking translations rather than the original
templates.
- Also fix return code not being passed on in check_files()
- Temporarily pin pip-tools, as its output format has changed
More than {{ is acceptable to start a tag, which means that `{{{Foo}}` won't be interpreted as "the content of `Foo`"
and should be rejected. For the sake of clarity and parsing, I suspect that those symbol should be rejected elsewhere
too.
Similary `{{Foo}}}` won't be interpreted as "Show the content of field `Foo}`" even if this field exists, so it's better
to reject `}`. It's clearly necessary to reject "}}" inside the field name, rejecting "}" seems easier to explain and
avoid future unexpected problem if the templates change.
The ":" are used to separate filters, and rejecting it in field name would ensure that there is no ambiguity.
