modpython: return a timestamp from the GetTime() variants #859

New Issue

2026-10-05T04:48:59+02:00

adam commented

2026-10-05 04:48:59 +02:00

Not sure if this affects modperl and others. Observed using the official (1.7) and git (master and 1.7.x) docker images.

Priority: low
Current result: <Swig Object of type 'timeval *' at 0x7f...>
Expected: some flavor of timestamp, such as...

A float: fractional seconds since the Epoch, as returned by Python's time.time() or Perl's Time::HiRes::time()
A two-tuple with secs and subsecs as ints, though the units (resolution/granularity) of the latter element would have to be documented somewhere
Some custom object with named members, thus avoiding the units question in (2)
A string: perhaps ISO-8601-style, like that passed via the time tag (millisecs)

For now, most users (like myself) are probably fine making do by converting "time" tags, if server-time is enabled, or otherwise calling time.time() (or similar) from within a hook.

Notes on (1) above...

Seems Python (On Unix) does the same thing as CUtils::GetTime() in going with nanosecs (timespec) from clock_gettime before microsecs (timeval) from gettimeofday.

The conversion is the usual scaling up to combine secs and subsecs (timespec version) and down again for the double. Guessing SWIG would take it from there and magically deliver whatever float is native to the target language. But if that's just nonsense, apologies.

Can provide example similar to the one in #1543, if necessary.

Not sure if this affects modperl and others. Observed using the official (1.7) and git (master and 1.7.x) docker images. Priority: low Current result: `<Swig Object of type 'timeval *' at 0x7f...>` Expected: some flavor of timestamp, such as... 1. A float: fractional seconds since the Epoch, as returned by Python's `time.time()` or Perl's `Time::HiRes::time()` 2. A two-tuple with secs and subsecs as ints, though the units (resolution/granularity) of the latter element would have to be documented somewhere 3. Some custom object with named members, thus avoiding the units question in (2) 4. A string: perhaps ISO-8601-style, like that passed via the [time tag][] (millisecs) For now, most users (like myself) are probably fine making do by converting "time" tags, if server-time is enabled, or otherwise calling `time.time()` (or similar) from within a hook. Notes on (1) above... Seems Python (On Unix) does the [same thing][] as `CUtils::GetTime()` in going with nanosecs (timespec) from `clock_gettime` before microsecs (timeval) from `gettimeofday`. The conversion is the usual scaling [up][] to combine secs and subsecs ([timespec version][]) and [down][] again for the double. Guessing SWIG would take it from there and magically deliver whatever float is native to the target language. But if that's just nonsense, apologies. Can provide example similar to the one in #1543, if necessary. [time tag]: https://ircv3.github.io/specs/extensions/server-time-3.2.html [same thing]: https://github.com/python/cpython/blob/16dfca4d829e45f36e71bf43f83226659ce49315/Python/pytime.c#L694 [up]: https://github.com/python/cpython/blob/16dfca4d829e45f36e71bf43f83226659ce49315/Python/pytime.c#L336 [timespec version]: https://github.com/python/cpython/blob/16dfca4d829e45f36e71bf43f83226659ce49315/Python/pytime.c#L290 [down]: https://github.com/python/cpython/blob/16dfca4d829e45f36e71bf43f83226659ce49315/Python/pytime.c#L447

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: adam/znc#859