fix bug: microsecond must be in 0..999999

[hujiaxing]

Note that issues in this repository are only for bugs or feature requests in the pywin32.

If you need support or help using this package, please follow these instructions - support or help requests will be closed without comment.

For all bugs, please provide the following information.

• Expected behavior and actual behavior.

• Steps to reproduce the problem.

• Version of Python and pywin32

[hujiaxing]

in PyTime.cpp

PyObject *PyWinObject_FromDATE(DATE t)
{

#ifdef PYWIN_HAVE_DATETIME_CAPI
// via https://www.codeproject.com/Articles/17576/SystemTime-to-VariantTime-with-Milliseconds
// (in particular, see the comments)
double fraction = t - (int)t; // extracts the fraction part
double hours = (fraction - (int)fraction) * 24.0;
double minutes = (hours - (int)hours) * 60.0;
double seconds = round((minutes - (int)minutes) * 60.0, 4);
double milliseconds = round((seconds - (int)seconds) * 1000.0, 0);
// the following is my fix
// if seconds has decimal part greater than 0.9995, it will be converted to 1000 milliseconds,
// and an exception will be caused: "microsecond must be in 0..999999",
// so set milliseconds to 999
if ((seconds - (int)seconds) >= 0.9995) {
milliseconds = 999.0;
}
// assert(milliseconds>=0.0 && milliseconds<=999.0);

    // Strip off the msec part of time
    double TimeWithoutMsecs = t - (ONETHOUSANDMILLISECONDS / 1000.0 * milliseconds);

    // Let the OS translate the variant date/time
    SYSTEMTIME st;
    if (!VariantTimeToSystemTime(TimeWithoutMsecs, &st)) {
        return PyWin_SetAPIError("VariantTimeToSystemTime");
    }
    if (milliseconds > 0.0) {
        // add the msec part to the systemtime object
        st.wMilliseconds = (WORD)milliseconds;
    }
    return PyWinObject_FromSYSTEMTIME(st);

#endif // PYWIN_HAVE_DATETIME_CAPI

#ifndef NO_PYWINTYPES_TIME
return new PyTime(t);
#endif
}

Is that a properly fix?

[mhammond]

Do you have a test case for the issue being fixed?

[hujiaxing]

Do you have a test case for the issue being fixed?

I used pywin32 to access outlook and sometimes exception thrown by pywin32.
It is critical value problem, so I just set milliseconds to 999 and program work fine.

[hujiaxing]

@mhammond

In PyTime.cpp

static double round(double Value, int Digits)
{
assert(Digits >= -4 && Digits <= 4);
int Idx = Digits + 4;
double v[] = {1e-4, 1e-3, 1e-2, 1e-1, 1, 10, 1e2, 1e3, 1e4};
return floor(Value * v[Idx] + 0.5) / (v[Idx]);
}

double milliseconds = round((seconds - (int)seconds) * 1000.0, 0); =>

double value = seconds - (int)seconds; // >= 0.9995
value = value * 1000; // >= 999.5
round(value, 0); =>

  return floor(value + 0.5) // will return 1000.0

so milliseconds will cause an exception

[mhammond]

FWIW, I spent some time trying to reproduce this with the test COM objects and failed. I'll have another go soon, but I'm reluctant to check in a speculative fix without any certainty around whether it actually fixes the problem.

[hujiaxing]

I have the same problem with the following link:
https://stackoverflow.com/questions/62169709/python-valueerror-microsecond-must-be-in-0-999999-while-using-win32com#

[hujiaxing]

Also this issue: #1466 Some date values in Excel cause ValueError: microsecond must be in 0..999999

[hujiaxing]

@mhammond any suggestion?

[mhammond]

@mhammond any suggestion?

The links above are other reports of the same bug, which I don't doubt exists. However, they don't seem to get me any closer to a repro.

[hujiaxing]

My fix is not good, I know that
I think there should be another patch for this bug
This bug has been exist for a long time
@mhammond

[mhammond]

My fix is not good,

It might be good!

This bug has been exist for a long time

Yeah - which is why I'd rather spend a little more time trying to get a repro rather than making a speculative fix. A tweak to the c++ test server that manages to find a float that causes this problem should be possible, but I failed to find it in my first attempt.

[hujiaxing]

@mhammond I construct a double date with milliseconds: 18712.308206013888

[bardal]

Is there a workaround? I am getting datetimes from Outlook (ReceivedTime on a Message). I don't need microseconds so would happily zero them out ahead of referencing the ReceivedTime which generates the ValueError Exception. Is that possible?

[bardal]

Alternatively, do you think this bugfix will be released in the near future?

[mhammond]

I just merged that PR, so it will be in the next release, but I can't predict when that might be.