Thread: Translatable strings with formatting of 64bit values

Translatable strings with formatting of 64bit values

From
Ildus Kurbangaliev
Date:
Hi,

apparently gettext can't properly identify strings when 64bit values
formatted with macros like INT64_FORMAT and UINT64_FORMAT. I did
some research and found out that gettext can work with PRId64 and
PRIu64. My suggestion is to use these macro for such strings.

The problem is here that PRIu64 is not accessible on all platforms but
this is easy solvable if it will be specified using INT64_MODIFIER in
c.h.

I attached a sample patch that adds PRIu64, PRId64 and makes few strings
translatable.

Using PRId64 will simplify the code like this:

        char        bufv[100],
                    bufm[100],
                    bufx[100];

        snprintf(bufv, sizeof(bufv), INT64_FORMAT, next);
        snprintf(bufm, sizeof(bufm), INT64_FORMAT, minv);
        snprintf(bufx, sizeof(bufx), INT64_FORMAT, maxv);
        ereport(ERROR,
                (errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
                 errmsg("setval: value %s is out of
        bounds for sequence \"%s\" (%s..%s)", bufv,
        RelationGetRelationName(seqrel), bufm, bufx)));


To:

    if ((next < minv) || (next > maxv))
        ereport(ERROR,
                (errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
                 errmsg("setval: value %s is out of
    bounds for sequence \"%" PRId64 "\" (%" PRId64 "..%" PRId64
    ")", next, RelationGetRelationName(seqrel), minv, maxv)));

In result:

#: commands/sequence.c:944
#, fuzzy, c-format
#| msgid "setval: value %s is out of bounds for sequence
\"%s\" (%s..%s)"
msgid "setval: value %s is out of bounds for sequence
\"%<PRId64>\" (%<PRId64>..%<PRId64>)"
msgstr "setval передано значение %s вне пределов последовательности
\"%s\" (%s..%s)"

And still this string will be translatable. I found a bunch of places
when PRIx64 macros can simplify the code.

--
---
Ildus Kurbangaliev
Postgres Professional: http://www.postgrespro.com
Russian Postgres Company

Attachment

Re: Translatable strings with formatting of 64bit values

From
Tom Lane
Date:
librelpIldus Kurbangaliev <i.kurbangaliev@postgrespro.ru> writes:
> apparently gettext can't properly identify strings when 64bit values
> formatted with macros like INT64_FORMAT and UINT64_FORMAT. I did
> some research and found out that gettext can work with PRId64 and
> PRIu64. My suggestion is to use these macro for such strings.

I do not believe what you're suggesting can actually work reliably.
In particular it'd fail if we're using our own version of snprintf
and that has different length modifiers than whatever gettext thinks
is the platform standard.  Also, how do you think it can possibly work
on platforms lacking the PRId64 macro?  gettext would have no better
idea than we do about what to assume that is.

Even if we could trust it to work, it doesn't seem to me that this:

>         ereport(ERROR,
>                 (errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
>                  errmsg("setval: value %s is out of
>     bounds for sequence \"%" PRId64 "\" (%" PRId64 "..%" PRId64
>     ")", next, RelationGetRelationName(seqrel), minv, maxv)));

is really any cleaner or easier to read than what we're doing now.

An idea that might be worth considering is to provide a helper function
that converts an int64 to a palloc'd string, so that you could write

    errmsg("setval: value %s is out of bounds ...",
            int64tostr(next), ...)

If you were executing in a long-lived memory context, this might risk
a memory leak, but we've not had a lot of problems with using other
functions that allocate some memory within ereport calls.  In any case
the existing method with a local buffer would work as a fallback
if that were a concern.

            regards, tom lane