date-time "l" output format code is unimplemented

[mhemmer53]

Boost 1.72.0, VS 2019, Windows 10, toolset v142 "l" is not recognized and appears unimplemented, attempted use causes abort to be called, no exception is thrown, suggested fix attached time_facet.zip

[JeffGarland]

It isn't documented to be either -- but it's a valid request.

https://www.boost.org/doc/libs/1_73_0/doc/html/date_time/date_time_io.html#date_time.format_flags

[JeffGarland]

is is possible to get and unzip upload -- wary of opening those for security reasons

[mhemmer53]

I see %l documented for output, I fact it causes my program to abort.

From: Jeff Garland [email protected] Sent: Monday, June 1, 2020 09:33 To: boostorg/date_time [email protected] Cc: Mike Hemmer [email protected]; Author [email protected] Subject: Re: [boostorg/date_time] date-time "l" output format code is unimplemented (#156)

It isn't documented to be either -- but it's a valid request.

https://www.boost.org/doc/libs/1_73_0/doc/html/date_time/date_time_io.html#date_time.format_flags

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/boostorg/date_time/issues/156#issuecomment-636894279, or unsubscribehttps://github.com/notifications/unsubscribe-auth/APYF3CN3WKP3V7DJAXECOQDRUO3YHANCNFSM4NNJG6NQ.

[mhemmer53]

Jeff,

The file I modified is attached.

Thanks,

Mike Hemmer

From: Jeff Garland [email protected] Sent: Monday, June 1, 2020 09:33 To: boostorg/date_time [email protected] Cc: Mike Hemmer [email protected]; Author [email protected] Subject: Re: [boostorg/date_time] date-time "l" output format code is unimplemented (#156)

is is possible to get and unzip upload -- wary of opening those for security reasons

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/boostorg/date_time/issues/156#issuecomment-636894649, or unsubscribehttps://github.com/notifications/unsubscribe-auth/APYF3CM7LDU5WJ3HZLAXMWLRUO323ANCNFSM4NNJG6NQ.

[k-ross]

It isn't documented to be either -- but it's a valid request.

https://www.boost.org/doc/libs/1_73_0/doc/html/date_time/date_time_io.html#date_time.format_flags

Actually it is documented, at the link you provided. Scroll down past the "date facet format flags" to "time facet format flags". It's listed there.

[JeffGarland]

Yes I missed it in the docs. If you look at the tests (testtime_input_facet.cpp) it's clear on the input side that it's not implemented (commented out test). On the output side, nominally the default behavior is to dispatch std::time_put for all the flags -- we only intercept stuff that extends normal c/c++ formatting. %I is covered in the base set for time_put, so it actually should work. So this maybe a VC problem and not a date_time problem.

[mhemmer53]

Jeff,

I do not see %l documented as a valid put_time format code here: https://docs.w3cub.com/cpp/io/manip/put_time/ Or here: https://en.cppreference.com/w/cpp/io/manip/put_time Or here: http://www.cplusplus.com/reference/iomanip/put_time/

Mike

From: Jeff Garland [email protected] Sent: Wednesday, June 3, 2020 19:36 To: boostorg/date_time [email protected] Cc: Mike Hemmer [email protected]; Author [email protected] Subject: Re: [boostorg/date_time] date-time "l" output format code is unimplemented (#156)

Yes I missed it in the docs. If you look at the tests (testtime_input_facet.cpp) it's clear on the input side that it's not implemented (commented out test). On the output side, nominally the default behavior is to dispatch std::time_put for all the flags -- we only intercept stuff that extends normal c/c++ formatting. %I is covered in the base set for time_put, so it actually should work. So this maybe a VC problem and not a date_time problem.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/boostorg/date_time/issues/156#issuecomment-638531452, or unsubscribehttps://github.com/notifications/unsubscribe-auth/APYF3CMRHCH7MYBWTIYHIBTRU3T7DANCNFSM4NNJG6NQ.

[JeffGarland]

Look again really close: https://en.cppreference.com/w/cpp/io/manip/put_time -- you'll see this:

I	writes hour as a decimal number, 12 hour clock (range [01,12])	tm_hour

I've confirmed the following works on linux with gcc9.2. I'll also note that there's a widestream test of this same code that passes on all the platforms. Note the wstream test is in an ifdef, but don't think windows would be excluded from it.

  ptime t(  date(2020,Oct,13), time_duration(18,01,56));
  time_facet* timefacet = new time_facet();
  timefacet->format("%Y-%b-%d %I:%M:%S %p");
  cout.imbue(std::locale(cout.getloc(), timefacet));
  cout << "%I test: " << t << endl;  //%I test: 2020-Oct-13 06:01:56 PM

Unfortunately I can't seem to bend godbolt to include boost for vc++

https://godbolt.org/z/bus8ft

[mhemmer53]

Jeff,

The format code I’m looking for is lowercase ell, not uppercase eye.

Mike

From: Jeff Garland [email protected] Sent: Thursday, June 4, 2020 08:45 To: boostorg/date_time [email protected] Cc: Mike Hemmer [email protected]; Author [email protected] Subject: Re: [boostorg/date_time] date-time "l" output format code is unimplemented (#156)

Look again really close: https://en.cppreference.com/w/cpp/io/manip/put_time -- you'll see this: I writes hour as a decimal number, 12 hour clock (range [01,12]) tm_hour

I've confirmed the following works on linux with gcc9.2. I'll also note that there's a widestream test of this same code that passes on all the platforms. Note the wstream test is in an ifdef, but don't think windows would be excluded from it.

ptime t( date(2020,Oct,13), time_duration(18,01,56));

time_facet* timefacet = new time_facet();

timefacet->format("%Y-%b-%d %I:%M:%S %p");

cout.imbue(std::locale(cout.getloc(), timefacet));

cout << "%I test: " << t << endl; //%I test: 2020-Oct-13 06:01:56 PM

Unfortunately I can't seem to bend godbolt to include boost for vc++

https://godbolt.org/z/bus8ft

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/boostorg/date_time/issues/156#issuecomment-638857418, or unsubscribehttps://github.com/notifications/unsubscribe-auth/APYF3CL6FJAZOISYMKFVA7DRU6QOHANCNFSM4NNJG6NQ.

[JeffGarland]

ok right -- really sorry for the confusion. So I looked it up. The time_put is specified in terms of iso/iec 9945 strptime which does not provide the '%l' (where that's an ell, not an eye). That said, some platforms are likely using strftime which does support. So this does work on Linux/gcc. Should be noted that c++20 std::chrono and std::format is still based on strptime, so this literal won't be available there.

https://godbolt.org/z/oKRcJZ