gh-158252: Extend C strptime timezone parsing

Allow nonterminal offsets, offset seconds and fractions, and %:z.
Preserve fractional offsets for datetime and time objects, and preserve
terminal UTC/GMT names and DST metadata for time.strptime.

Local timezone names and nonterminal %Z still use the Python parser.

Author: pganssle

Include/internal/pycore_time.h

@@ -348,6 +348,8 @@ typedef struct {
     int year, month, day;
     int hour, minute, second, fraction;
     int gmtoff;  // INT_MIN means no offset was supplied.
+    int gmtoff_fraction;
+    Py_ssize_t zone_start, zone_length;  // Original spelling of a %Z name.
     int yday;    // January 1 is day 1.
     int weekday;  // Monday is 0; a supplied weekday need not agree with the date.
 } _PyTime_StrptimeFields;

Lib/test/test_strptime.py

@@ -16,6 +16,21 @@
 
 
 class PublicStrptimeTests(unittest.TestCase):
+    def test_offsets_and_zone_names(self):
+        offsets = ('', 'Z', '+0530', '-03:30', '+01:02:03', '-000001.000001',
+                   '+01:02:03.456789', '+000000.5', '-00:00:00.5',
+                   '+2400', '+0060', '+01:0203', '+0102:03', '+01:02:99',
+                   '+01:02:03.', '+01:02:03.1234567', 'z', '+', '+000')
+        for offset in offsets:
+            for directive in ('%z', '%:z'):
+                for suffix, fmt_suffix in (('', ''), (' UTC', ' %Z'),
+                                           (' gMt', ' %Z'), (' 12:34', ' %H:%M')):
+                    self.assert_matches_python(offset + suffix,
+                                               directive + fmt_suffix)
+        for name in ('UTC', 'GMT', 'uTc', 'gmT', 'UTCx', 'LOCAL', ''):
+            self.assert_matches_python(name, '%Z')
+        self.assert_matches_python('+0100 +0200', '%z %:z')
+
     @run_with_locales('LC_TIME', 'C', 'en_US', 'fr_FR', 'de_DE', 'ja_JP', '')
     def test_twelve_hour_clock(self):
         for hour in range(24):

Modules/_datetimemodule.c

@@ -1685,6 +1685,27 @@ tzinfo_from_isoformat_results(int rv, int tzoffset, int tz_useconds)
  * String format helpers.
  */
 
+static PyObject *
+tzinfo_from_strptime_results(const _PyTime_StrptimeFields *fields,
+                             PyObject *string)
+{
+    if (!fields->zone_length || fields->gmtoff == INT_MIN) {
+        return tzinfo_from_isoformat_results(fields->gmtoff != INT_MIN,
+                                            fields->gmtoff,
+                                            fields->gmtoff_fraction);
+    }
+    PyObject *name = PyUnicode_Substring(string, fields->zone_start,
+                                        fields->zone_start + fields->zone_length);
+    if (name == NULL) {
+        return NULL;
+    }
+    PyObject *delta = new_delta(0, fields->gmtoff, fields->gmtoff_fraction, 1);
+    PyObject *tzinfo = delta == NULL ? NULL : new_timezone(delta, name);
+    Py_XDECREF(delta);
+    Py_DECREF(name);
+    return tzinfo;
+}
+
 static PyObject *
 format_ctime(PyObject *date, int hours, int minutes, int seconds)
 {
@@ -4782,8 +4803,7 @@ datetime_time_strptime_impl(PyTypeObject *type, PyObject *string,
     _PyTime_StrptimeFields fields;
     if (type == TIME_TYPE(NO_STATE) &&
         _PyTime_Strptime(string, format, &fields)) {
-        PyObject *tzinfo = tzinfo_from_isoformat_results(
-            fields.gmtoff != INT_MIN, fields.gmtoff, 0);
+        PyObject *tzinfo = tzinfo_from_strptime_results(&fields, string);
         if (tzinfo == NULL) {
             return NULL;
         }
@@ -5877,8 +5897,7 @@ datetime_datetime_strptime_impl(PyTypeObject *type, PyObject *string,
     _PyTime_StrptimeFields fields;
     if (type == DATETIME_TYPE(NO_STATE) &&
         _PyTime_Strptime(string, format, &fields)) {
-        PyObject *tzinfo = tzinfo_from_isoformat_results(
-            fields.gmtoff != INT_MIN, fields.gmtoff, 0);
+        PyObject *tzinfo = tzinfo_from_strptime_results(&fields, string);
         if (tzinfo == NULL) {
             return NULL;
         }

Modules/timemodule.c

@@ -1019,7 +1019,7 @@ time_strptime(PyObject *self, PyObject *args)
         const int values[] = {
             fields.year, fields.month, fields.day,
             fields.hour, fields.minute, fields.second,
-            fields.weekday, fields.yday, -1
+            fields.weekday, fields.yday, fields.zone_length ? 0 : -1
         };
         for (size_t i = 0; i < Py_ARRAY_LENGTH(values); i++) {
             PyObject *value = PyLong_FromLong(values[i]);
@@ -1029,8 +1029,15 @@ time_strptime(PyObject *self, PyObject *args)
             }
             PyStructSequence_SET_ITEM(result, i, value);
         }
-        /* Numeric offsets don't supply a timezone name or DST information. */
-        PyStructSequence_SET_ITEM(result, 9, Py_NewRef(Py_None));
+        PyObject *zone = fields.zone_length
+            ? PyUnicode_Substring(PyTuple_GET_ITEM(args, 0), fields.zone_start,
+                                  fields.zone_start + fields.zone_length)
+            : Py_NewRef(Py_None);
+        if (zone == NULL) {
+            Py_DECREF(result);
+            return NULL;
+        }
+        PyStructSequence_SET_ITEM(result, 9, zone);
         PyObject *offset = fields.gmtoff == INT_MIN
                           ? Py_NewRef(Py_None)
                           : PyLong_FromLong(fields.gmtoff);

Python/_strptime.c

@@ -11,11 +11,11 @@
  * - numeric %Y, %y, %m, %d, %H, %M, %S, and %f fields;
  * - ASCII locale month/weekday names and numeric %w/%u weekdays;
  * - %I with ASCII %p indicators, and the %k, %l, %P aliases;
- * - a terminal %z: empty, Z, or +/-HHMM or +/-HH:MM with hours below 24;
+ * - %z and %:z offsets with hours below 24, and terminal UTC/GMT %Z names;
  * - literal characters, %%, and ASCII whitespace.
  * Duplicate directives, mixed %Y/%y, and day-of-month without a year use the
  * fallback. So do other directives (including non-ASCII locale names and
- * week/day-of-year calculations), and offsets with seconds or fractions.
+ * week/day-of-year calculations).
  */
 
 static int
@@ -104,32 +104,58 @@ strptime_name(const unsigned char *data, Py_ssize_t length, Py_ssize_t pos,
 
 static int
 strptime_offset(const unsigned char *data, Py_ssize_t length, Py_ssize_t pos,
-                _PyTime_StrptimeFields *fields)
+                int require_colon, _PyTime_StrptimeFields *fields)
 {
-    Py_ssize_t remaining = length - pos;
-    if (remaining == 0) {
-        return 1;
+    Py_ssize_t start = pos;
+    if (pos == length) {
+        return 0;
     }
-    if (remaining == 1 && data[pos] == 'Z') {
+    if (data[pos] == 'Z') {
         fields->gmtoff = 0;
         return 1;
     }
-    /* Offsets containing seconds or fractions fall back to _strptime.py. */
-    if ((remaining != 5 && remaining != 6) ||
-        (data[pos] != '+' && data[pos] != '-')) {
+    if (data[pos] != '+' && data[pos] != '-') {
         return 0;
     }
-    int hours, minutes;
-    int colon = remaining == 6;
-    if ((colon && data[pos + 3] != ':') ||
-        strptime_digits(data, length, pos + 1, 2, 2, &hours) != 2 ||
-        strptime_digits(data, length, pos + 3 + colon, 2, 2, &minutes) != 2 ||
-        hours > 23 || minutes > 59) {
-        return 0;
+    int sign = data[pos++] == '-' ? -1 : 1;
+    int hours, minutes, seconds = 0, fraction = 0;
+    if (strptime_digits(data, length, pos, 2, 2, &hours) != 2 || hours > 23) {
+        return -1;
     }
-    fields->gmtoff = (hours * 3600 + minutes * 60) *
-                    (data[pos] == '-' ? -1 : 1);
-    return 1;
+    pos += 2;
+    int colon = pos < length && data[pos] == ':';
+    if (require_colon && !colon) {
+        return -1;
+    }
+    pos += colon;
+    if (strptime_digits(data, length, pos, 2, 2, &minutes) != 2 || minutes > 59) {
+        return -1;
+    }
+    pos += 2;
+    int seconds_colon = pos < length && data[pos] == ':';
+    Py_ssize_t seconds_pos = pos + seconds_colon;
+    if (strptime_digits(data, length, seconds_pos, 2, 2, &seconds) == 2) {
+        if (seconds > 59 || seconds_colon != colon) {
+            return -1;
+        }
+        pos = seconds_pos + 2;
+        if (pos < length && data[pos] == '.') {
+            int count = strptime_digits(data, length, pos + 1, 1, 6, &fraction);
+            if (count == 0) {
+                return -1;
+            }
+            pos += count + 1;
+            for (; count < 6; count++) {
+                fraction *= 10;
+            }
+        }
+    }
+    else {
+        seconds = 0;
+    }
+    fields->gmtoff = sign * (hours * 3600 + minutes * 60 + seconds);
+    fields->gmtoff_fraction = sign * fraction;
+    return (int)(pos - start);
 }
 
 /* Return 1 for a complete numeric match, or 0 if fallback is required.
@@ -156,6 +182,7 @@ _PyTime_Strptime(PyObject *string, PyObject *format, _PyTime_StrptimeFields *fie
         OFFSET = 1U << 8,
         WEEKDAY = 1U << 9,
         AMPM = 1U << 10,
+        ZONE = 1U << 11,
         ANY_YEAR = YEAR | SHORT_YEAR,
     };
     static const unsigned short directive_bits[128] = {
@@ -170,6 +197,7 @@ _PyTime_Strptime(PyObject *string, PyObject *format, _PyTime_StrptimeFields *fie
         ['z'] = OFFSET,
         ['w'] = WEEKDAY,
         ['p'] = AMPM,
+        ['Z'] = ZONE,
     };
     /* String subclasses and non-ASCII input retain Python's behavior. */
     if (!PyUnicode_CheckExact(string) || !PyUnicode_CheckExact(format) ||
@@ -211,6 +239,11 @@ _PyTime_Strptime(PyObject *string, PyObject *format, _PyTime_StrptimeFields *fie
             return 0;
         }
         c = fmt[i];
+        int require_colon = 0;
+        if (c == ':' && i + 1 < fmt_length && fmt[i + 1] == 'z') {
+            c = fmt[++i];
+            require_colon = 1;
+        }
         if (c == 'k') {
             c = 'H';
         }
@@ -266,14 +299,32 @@ _PyTime_Strptime(PyObject *string, PyObject *format, _PyTime_StrptimeFields *fie
             continue;
         }
 
-        if (c == 'z') {
-            if (i != fmt_length - 1 ||
-                !strptime_offset(data, length, pos, fields)) {
+        if (c == 'Z') {
+            /* Local timezone names and nonterminal names retain Python's
+             * tzset state, DST rules and regex backtracking.
+             */
+            if (i != fmt_length - 1 || length - pos != 3 ||
+                !((Py_TOLOWER(data[pos]) == 'u' &&
+                   Py_TOLOWER(data[pos + 1]) == 't' &&
+                   Py_TOLOWER(data[pos + 2]) == 'c') ||
+                  (Py_TOLOWER(data[pos]) == 'g' &&
+                   Py_TOLOWER(data[pos + 1]) == 'm' &&
+                   Py_TOLOWER(data[pos + 2]) == 't'))) {
                 return 0;
             }
+            fields->zone_start = pos;
+            fields->zone_length = 3;
             pos = length;
             continue;
         }
+        if (c == 'z') {
+            int count = strptime_offset(data, length, pos, require_colon, fields);
+            if (count < 0) {
+                return 0;
+            }
+            pos += count;
+            continue;
+        }
 
         int minimum = 1;
         int maximum = 2;