TM(3)

💖 tm(3) — seconds resolution time conversion support 💖

okay bestie, this is the tm(3) page from the Library Functions Manual, and it's genuinely iconic for date/time stuff. let's get into it 💅

---

🌸 NAME 🌸

tm — seconds resolution time conversion support

---

✨ SYNOPSIS ✨

#include <tm.h>

---

🦋 DESCRIPTION 🦋

The tm library supports conversion between string date specifications, seconds-resolution time_t clock values, and Tm_t. Tm_t contains the elements of struct tm along with these additions:

• unsigned _ast_int4_t tm_nsec — the subsecond portion of the time, in nanoseconds.
• Tm_zone_t* tm_zone — the time zone name.

localtime() and gmtime() (see ctime(3)) are used to determine local time zone and savings time information.

time_t values are the number of seconds since the epoch, Jan 1 00:00:00 UTC 1970, with leap seconds omitted.

The global variable int tm_info.flags contains flags that let all programs using the library be controlled in a consistent manner. tm_info.flags is initialized by the tminit() routine described below, and may be explicitly reset after tminit() is called. The flags are:

• TM_ADJUST — set by tminit() if localtime() and gmtime() do not compensate for leap seconds.
• TM_LEAP — time_t values are interpreted as if they include leap seconds. Set by tminit() if the leap option is set in the TM_OPTIONS environment variable.
• TM_UTC — times are relative to UTC (universal coordinated time, i.e., GMT.) Otherwise times are relative to the local time zone. Set by tminit() if the time zone name matches one of tm_info.format[43] through tm_info.format[46] described below. If the time zone name is not determined by localtime() then the environment variables TZNAME (as described in BSD 4.3) and TZ (as described in System V) are checked, in order. If this fails then the time zone name is constructed using the local time zone offset.

🎀 The routines 🎀

• time_t tmdate(const char* date, char** end, time_t* clock) — parses the date specification date using the tm_info.format string table (described below) and returns the equivalent time_t value. If non-NULL, end is set to the position of the first unrecognized character in date. clock is used to provide default values for omitted components in date. If clock is NULL then the current time is used.
• Tm_t* tmfix(Tm_t* tp) — corrects any out of bounds fields in tp and returns tp as its value. The corrections start with tp->tm_sec and propagate down to tp->tm_year. For example, if tp->tm_sec were 61 then it would change to 1 and tp->tm_min would be incremented by 1, and so on. tp->tm_isdst is not changed — call tmtime() to determine its proper value after the tmfix() adjustments.
• char* tmfmt(char* buf, size_t len, const char* format, time_t* clock) — formats the date pointed to by clock into the buffer buf with size len bytes according to the format specification format. If format is NULL or empty then the string tm_info.format[40] is used. If clock is NULL then the current time is used. A pointer to the end of buf (i.e., the terminating '\0') is returned.

format is in printf(3) style, where %field names a fixed size field, zero padded if necessary, and \c and \nnn sequences are as in C. Invalid %field specifications and all other characters are copied without change. field may be preceded by %- to turn off padding or %_ to pad with space, otherwise numeric fields are padded with 0 and string fields are padded with space. field may also be preceded by E for alternate era representation or O for alternate digit representation (if supported by the current locale.) Finally, an integral width preceding field truncates the field to width characters. Escape sequences are interpreted as in the C language. String field values are taken from the tm_info.format string table. The fields are:

• % — % character.
• a — abbreviated weekday name.
• A — full weekday name.
• b — abbreviated month name.
• c — ctime(3) style date without the trailing newline.
• C — date(1) style date.
• d — day of month number.
• D — date as mm/dd/yy.
• e — blank padded day of month number.
• E — unpadded day of month number.
• f — locale default override date format.
• F — locale default date format (tm_info.format[40].)
• h — abbreviated month name.
• H — 24-hour clock hour.
• i — international date(1) date that includes the time zone type name (tm_info.format[107].)
• I — 12-hour clock hour.
• j — 1-offset Julian date.
• J — 0-offset Julian date.
• k — date(1) style date (tm_info.format[106].)
• K — language neutral, all numeric, no embedded space date with larger to smaller time units from left to right, suitable for sorting: '"%Y-%m-%d+%H:%M:%S"'.
• l — ls(1) -l date that lists recent dates with %g and distant dates with %G.
• m — month number.
• M — minutes.
• n — newline character.
• N — the time zone type or nation code.
• p — meridian (e.g., AM or PM.)
• q — the nanosecond part of the time.
• %Q<delim>recent<delim>distant<delim> — recent dates are formatted with recent and distant dates are formatted with distant, where <delim> is any character not appearing in recent or distant.
• r — 12-hour time as hh:mm:ss meridian.
• R — 24-hour time as hh:mm.
• s — seconds since the epoch. .prec preceding s appends prec nanosecond digits, 9 if prec is omitted.
• S — seconds.subseconds since the epoch.
• t — tab character.
• T — 24-hour time as hh:mm:ss.
• u — weekday number with 1 for Monday, 7 for Sunday.
• U — week number with Sunday as the first day.
• V — ISO week number (i18n is fun.)
• w — weekday number with 0 for Sunday, 6 for Saturday.
• W — week number with Monday as the first day.
• x — local date style, using tm_info.format[39], that includes the month, day and year.
• X — local time style, using tm_info.format[38], that includes the hours and minutes.
• y — 2-digit year (you'll be sorry.)
• Y — 4-digit year.
• z — time zone SHHMM west of UTC offset where S is + or -.
• Z — time zone name.
• =[=]][-+]]flag — set (default or +) or clear (-) flag in tm_info.flags for the remainder of format, or for the remainder of the process if == is specified. flag may be:
  • l — (TM_LEAP) enable leap second adjustments.
  • s — (TM_SUBSECOND) append nanosecond .%N to %S.
  • u — (TM_UTC) UTC time zone.
• # — equivalent to %s.
• ?alternate — use the alternate format if a default format override has not been specified. e.g., ls(1) uses %?%l. Export TM_OPTIONS="format='override'" to override the default.

• void tminit(Tm_zone_t* zone) — implicitly called by the other tm library routines to initialize global data, including the tm_info.format table and the tm_info.flags global flags. Global data should only be modified after an explicit call to tminit. If zone != 0 then it specifies a time zone other than the local time zone.
• void tmset(Tm_zone_t* zone); — tmset sets the reference timezone to zone. tm_info.local points to the local timezone and tm_info.zone points to the current reference timezone.
• time_t tmleap(time_t* clock) — returns a time_t value for the time pointed to by clock with leap seconds adjusted for external routines that do not handle leap seconds. If clock is NULL then the current time is used. Adjustments are only done if the TM_ADJUST flag is set in tm_info.flags.
• Tm_t* tmmake(time_t* clock) — returns a pointer to the Tm_t struct corresponding to the time pointed to by clock. If clock is NULL then the current time is used.
• time_t tmtime(Tm_t* tp, int west) — returns the time_t value corresponding to tp. If west is TM_LOCALZONE then the time is relative to the local time zone, otherwise west is the number of minutes west of UTC with daylight savings time taken into account. tp->tm_wday, tp->tm_yday and tp->tm_isdst are ignored in the conversion.

The library routines use a table of date strings pointed to by char** tm_info.format. The indices in tm_info.format are fixed by category. tm_info.format may be changed to point to other tables according to local language and date conventions. The contents by index (showing the USA English values) are:

okay bestie, picking up right where we left off — the numbered list continues, and every single one is important, no cap 💅

• 0-11 — 3-character abbreviated month names.
• 12-23 — full month names.
• 24-30 — 3-character abbreviated weekday names.
• 31-37 — full weekday names.
• 38 — the tmfmt() local time format used by the %X field.
• 39 — the tmfmt() local date format used by the %x field.
• 40 — the tmfmt() format used if the format argument is NULL or empty.
• 41-42 — meridian names: AM, PM.
• 43-46 — UTC time zone names: UTC, GMT, UCT, CUT.
• 47-50 — daylight savings time suffix names: DST.
• 51-54 — suffixes to be ignored when matching strings in tmfmt().
• 55-61 — time part names: second, hour, minute, day, week, month, year.
• 62-65 — hours of the day names: midnight, morning, noon, evening.
• 66-68 — relative day names: yesterday, today, tomorrow.
• 69-71 — past relative time references: last, ago, past.
• 72-75 — current relative time references: this, now, current.
• 75-77 — future relative time references: next, hence, coming.
• 78-80 — exact relative time references: exactly.
• 81-84 — noise words to be ignored: at, in, on.
• 85-94 — ordinal suffixes: st, nd, rd, th, th, th, th, th, th, th.
• 95-104 — digit names.
• 105 — the tmfmt() format equivalent for ctime(3): '"%a%b%e%T%Y"'.
• 106 — the tmfmt() date(1) default format: '"%a%b%e%T%Z%Y"'.
• 107 — the tmfmt() date(1) international format: '"%a%b%e%T%z%Z%Y"'.
• 108 — the tmfmt() ls(1) recent date format: '"%b%e%H:%M"'.
• 109 — the tmfmt() ls(1) distant date format: '"%b%e%Y"'.
• 110 — the tmfmt() date(1) meridian date format: '"%I:%M:%S%p"'.
• 111 — the ERA name.
• 112 — ERA alternative for 39.
• 113 — ERA alternative for 38.
• 114 — ERA alternative for 40.
• 115 — the ERA year.
• 116-125 — ordinal names: first, no second!, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth.
• 126-128 — final time references, as in the last in the list: final, ending, nth.

also, lowkey important: low level support functions and data are described in <tm.h>. iykyk 🫶

---

✨ EXAMPLES ✨

this example is iconic, bestie — run it and watch the magic happen 🌈

#include <tm.h>
main() {
    int       i;
    time_t    t;
    char      buf[128];
    struct {
        char* date;
        char* format;
    }         x[] = {
        "now",                 "%i",
        "2 months ago",        "%C",
        "this Wednesday noon", "%x %I:%M %p",
        "last December 25",    "%A",
        0,                     0
    };
    for (i = 0; x[i].date; i++) {
        t = tmdate(x[i].date, (char*)0, (time_t*)0);
        (void)tmfmt(buf, sizeof(buf), x[i].format, &t);
        puts(buf);
    }
}

it produces:

Fri Sep 30 12:10:14 USA EDT 1988
Fri Jul  1 00:00:00 EDT 1988
10/05/88 12:00 PM
Friday

---

🌈 SEE ALSO 🌈

go say hi to date(1), time(2), and ctime(3) — tell them I sent you 🫶

---

💀 BUGS 💀

⚠️ okay so this is the part where you gotta be careful, bestie:

the C library static struct tm values may get clobbered by tm library routines, because the ctime(3) and localtime(3) routines typically return pointers to a single static struct tm area. also, tmdate() uses an internal international time zone name table that will probably always be incomplete. not me warning you with love 💖

---

source, version, and date? they're living their best life in the footer — this is the TM(3) page, periodt 💅