[libc] Implement strftime

TODO: DESCRIPTION
TODO: TESTS
TODO: RotFO (Rest of the Formatting Options)

Roughly based on llvm#111305, but with major rewrites.

Author: michaelrj-google

libc/config/linux/x86_64/entrypoints.txt

@@ -1103,6 +1103,7 @@ if(LLVM_LIBC_FULL_BUILD)
     libc.src.time.gmtime_r
     libc.src.time.mktime
     libc.src.time.nanosleep
+    libc.src.time.strftime
     libc.src.time.time
     libc.src.time.timespec_get
 

libc/docs/dev/undefined_behavior.rst

@@ -106,3 +106,27 @@ uninitialized spinlock and invalid spinlock is left undefined. We follow the rec
 POSIX.1-2024, where EINVAL is returned if the spinlock is invalid (here we only check for null pointers) or
 EBUSY is returned if the spinlock is currently locked. The lock is poisoned after a successful destroy. That is,
 subsequent operations on the lock object without any reinitialization will return EINVAL.
+
+Strftime
+--------
+In the C Standard, it provides a list of modifiers, and the conversions these
+are valid on. It also says that a modifier on an unspecified conversion is
+undefined. For LLVM-libc, the conversion is treated as if the modifier isn't
+there.
+
+If a struct tm with values out of the normal range is passed, the standard says
+the result is undefined. For LLVM-libc, the result may be either the normalized
+value (e.g. weekday % 7) or the actual, out of range value. This behavior is not
+necessarily consistent between conversions, even similar ones. For conversions
+that result in strings, passing an out of range value will result in "?".
+
+Posix adds padding support to strftime, but says "the default padding character
+is unspecified." For LLVM-libc, the default padding character is ' ' (space),
+for all string-type conversions and '0' for integer-type conversions.
+
+Posix also adds flags and a minimum field width, but leaves unspecified what
+happens for most combinations of these. For LLVM-libc:
+An unspecified minimum field width defaults to 0.
+More specific flags take precedence over less specific flags (i.e. '+' takes precedence over '0')
+Any conversion with a minimum width is padded with the padding character until it is at least as long as the minimum width.
+Modifiers are applied, then the result is padded if necessary.

libc/include/llvm-libc-types/struct_tm.h

@@ -19,6 +19,7 @@ struct tm {
   int tm_wday;  // days since Sunday
   int tm_yday;  // days since January
   int tm_isdst; // Daylight Saving Time flag
+  // TODO: add tm_gmtoff and tm_zone? (posix extensions)
 };
 
 #endif // LLVM_LIBC_TYPES_STRUCT_TM_H

libc/include/time.yaml

@@ -91,6 +91,25 @@ functions:
     arguments:
       - type: const struct timespec *
       - type: struct timespec *
+  - name: strftime
+    standard:
+      - stdc
+    return_type: size_t
+    arguments:
+      - type: char *__restrict
+      - type: size_t
+      - type: const char *__restrict
+      - type: const struct tm *__restrict
+  - name: strftime_l
+    standard:
+      - stdc
+    return_type: size_t
+    arguments:
+      - type: char *__restrict
+      - type: size_t
+      - type: const char *__restrict
+      - type: const struct tm *__restrict
+      - type: locale_t
   - name: time
     standard:
       - stdc

libc/src/time/CMakeLists.txt

@@ -22,6 +22,8 @@ add_object_library(
   DEPENDS
     libc.include.time
     libc.src.__support.CPP.limits
+    libc.src.__support.CPP.string_view
+    libc.src.__support.CPP.optional
     libc.src.errno.errno
     .time_constants
     libc.hdr.types.time_t
@@ -133,6 +135,21 @@ add_entrypoint_object(
     libc.hdr.types.struct_tm
 )
 
+add_subdirectory(strftime_core) #TODO: Move to top
+
+add_entrypoint_object(
+  strftime
+  SRCS
+    strftime.cpp
+  HDRS
+    strftime.h
+  DEPENDS
+    libc.hdr.types.size_t
+    libc.hdr.types.struct_tm
+    libc.src.stdio.printf_core.writer
+    libc.src.time.strftime_core.strftime_main
+)
+
 add_entrypoint_object(
   time
   SRCS

libc/src/time/mktime.cpp

@@ -14,16 +14,6 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-// Returns number of years from (1, year).
-static constexpr int64_t get_num_of_leap_years_before(int64_t year) {
-  return (year / 4) - (year / 100) + (year / 400);
-}
-
-// Returns True if year is a leap year.
-static constexpr bool is_leap_year(const int64_t year) {
-  return (((year) % 4) == 0 && (((year) % 100) != 0 || ((year) % 400) == 0));
-}
-
 LLVM_LIBC_FUNCTION(time_t, mktime, (struct tm * tm_out)) {
   // Unlike most C Library functions, mktime doesn't just die on bad input.
   // TODO(rtenneti); Handle leap seconds.
@@ -69,7 +59,7 @@ LLVM_LIBC_FUNCTION(time_t, mktime, (struct tm * tm_out)) {
     }
     tm_year_from_base += years;
   }
-  bool tm_year_is_leap = is_leap_year(tm_year_from_base);
+  bool tm_year_is_leap = time_utils::is_leap_year(tm_year_from_base);
 
   // Calculate total number of days based on the month and the day (tm_mday).
   int64_t total_days = tm_out->tm_mday - 1;
@@ -83,21 +73,25 @@ LLVM_LIBC_FUNCTION(time_t, mktime, (struct tm * tm_out)) {
   total_days += (tm_year_from_base - time_constants::EPOCH_YEAR) *
                 time_constants::DAYS_PER_NON_LEAP_YEAR;
   if (tm_year_from_base >= time_constants::EPOCH_YEAR) {
-    total_days += get_num_of_leap_years_before(tm_year_from_base - 1) -
-                  get_num_of_leap_years_before(time_constants::EPOCH_YEAR);
+    total_days +=
+        time_utils::get_num_of_leap_years_before(tm_year_from_base - 1) -
+        time_utils::get_num_of_leap_years_before(time_constants::EPOCH_YEAR);
   } else if (tm_year_from_base >= 1) {
-    total_days -= get_num_of_leap_years_before(time_constants::EPOCH_YEAR) -
-                  get_num_of_leap_years_before(tm_year_from_base - 1);
+    total_days -=
+        time_utils::get_num_of_leap_years_before(time_constants::EPOCH_YEAR) -
+        time_utils::get_num_of_leap_years_before(tm_year_from_base - 1);
   } else {
     // Calculate number of leap years until 0th year.
-    total_days -= get_num_of_leap_years_before(time_constants::EPOCH_YEAR) -
-                  get_num_of_leap_years_before(0);
+    total_days -=
+        time_utils::get_num_of_leap_years_before(time_constants::EPOCH_YEAR) -
+        time_utils::get_num_of_leap_years_before(0);
     if (tm_year_from_base <= 0) {
       total_days -= 1; // Subtract 1 for 0th year.
       // Calculate number of leap years until -1 year
       if (tm_year_from_base < 0) {
-        total_days -= get_num_of_leap_years_before(-tm_year_from_base) -
-                      get_num_of_leap_years_before(1);
+        total_days -=
+            time_utils::get_num_of_leap_years_before(-tm_year_from_base) -
+            time_utils::get_num_of_leap_years_before(1);
       }
     }
   }

libc/src/time/strftime.cpp

@@ -0,0 +1,31 @@
+//===-- Implementation of strftime function -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/time/strftime.h"
+#include "hdr/types/size_t.h"
+#include "hdr/types/struct_tm.h"
+#include "src/__support/common.h"
+#include "src/__support/macros/config.h"
+#include "src/stdio/printf_core/writer.h"
+#include "src/time/strftime_core/strftime_main.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+LLVM_LIBC_FUNCTION(size_t, strftime,
+                   (char *__restrict buffer, size_t buffsz,
+                    const char *__restrict format, const struct tm *timeptr)) {
+
+  printf_core::WriteBuffer wb(buffer, (buffsz > 0 ? buffsz - 1 : 0));
+  printf_core::Writer writer(&wb);
+  int ret = strftime_core::strftime_main(&writer, format, timeptr);
+  if (buffsz > 0) // if the buffsz is 0 the buffer may be a null pointer.
+    wb.buff[wb.buff_cur] = '\0';
+  return (ret < 0 || static_cast<size_t>(ret) > buffsz) ? 0 : ret;
+}
+
+} // namespace LIBC_NAMESPACE_DECL

libc/src/time/strftime.h

@@ -0,0 +1,23 @@
+//===-- Implementation header of strftime -----------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_TIME_STRFTIME_H
+#define LLVM_LIBC_SRC_TIME_STRFTIME_H
+
+#include "hdr/types/size_t.h"
+#include "hdr/types/struct_tm.h"
+#include "src/__support/macros/config.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+size_t strftime(char *__restrict, size_t max, const char *__restrict format,
+                const struct tm *timeptr);
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC_TIME_STRFTIME_H

libc/src/time/strftime_core/CMakeLists.txt

@@ -0,0 +1,51 @@
+add_header_library(
+  core_structs
+  HDRS
+    core_structs.h
+  DEPENDS
+    libc.src.__support.CPP.string_view
+    libc.hdr.types.struct_tm
+)
+
+add_header_library(
+  parser
+  HDRS
+    parser.h
+  DEPENDS
+    .core_structs
+    libc.src.__support.CPP.string_view
+    libc.src.__support.ctype_utils
+    libc.src.__support.str_to_integer
+)
+
+add_object_library(
+  converter
+  SRCS
+    converter.cpp
+  HDRS
+    converter.h
+    num_converter.h
+    # str_converter.h
+    # composite_converter.h
+  DEPENDS
+    .core_structs
+    libc.src.time.time_utils
+    libc.src.time.time_constants
+    libc.src.stdio.printf_core.writer
+    libc.src.__support.CPP.string_view
+    libc.src.__support.integer_to_string
+)
+
+add_object_library(
+  strftime_main
+  SRCS
+    strftime_main.cpp
+  HDRS
+    strftime_main.h
+  DEPENDS
+    .core_structs
+    .parser
+    .converter
+    libc.src.stdio.printf_core.writer
+    libc.hdr.types.struct_tm
+)

libc/src/time/strftime_core/converter.cpp

@@ -0,0 +1,91 @@
+//===-- Format specifier converter implmentation for strftime -------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See htto_conv.times://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/__support/macros/config.h"
+#include "src/stdio/printf_core/writer.h"
+#include "src/time/strftime_core/core_structs.h"
+
+// #include "composite_converter.h"
+#include "num_converter.h"
+// #include "str_converter.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace strftime_core {
+
+int convert(printf_core::Writer *writer, const FormatSection &to_conv,
+            const tm *timeptr) {
+  // TODO: Implement the locale support.
+  if (to_conv.modifier != ConvModifier::none) {
+    return writer->write(to_conv.raw_string);
+  }
+
+  if (!to_conv.has_conv)
+    return writer->write(to_conv.raw_string);
+  switch (to_conv.conv_name) {
+    // The cases are grouped by type, then alphabetized with lowercase before
+    // uppercase.
+
+    // raw conversions
+  case '%':
+    return writer->write("%");
+  case 'n':
+    return writer->write("\n");
+  case 't':
+    return writer->write("\t");
+
+    // numeric conversions
+  case 'C': // Century
+  case 'd': // Day of the month [01-31]
+  case 'e': // Day of the month [1-31]
+  case 'g': // last 2 digits of ISO year
+  case 'G': // ISO year
+  case 'H': // 24-hour format
+  case 'I': // 12-hour format
+  case 'j': // Day of the year
+  case 'm': // Month of the year
+  case 'M': // Minute of the hour
+  case 's': // Seconds since the epoch
+  case 'S': // Second of the minute
+  case 'u': // ISO day of the week ([1-7] starting Monday)
+  case 'U': // Week of the year ([00-53] week 1 starts on first *Sunday*)
+  case 'V': // ISO week number ([01-53], 01 is first week majority in this year)
+  case 'w': // Day of week ([0-6] starting Sunday)
+  case 'W': // Week of the year ([00-53] week 1 starts on first *Monday*)
+  case 'y': // Year of the Century
+  case 'Y': // Full year
+    return convert_int(writer, to_conv, timeptr);
+
+    // string conversions
+  case 'a': // Abbreviated weekday name
+  case 'A': // Full weekday name
+  case 'b': // Abbreviated month name
+  case 'B': // Full month name
+  case 'h': // same as %b
+  case 'p': // AM/PM designation
+  case 'z': // Timezone offset (+/-hhmm)
+  case 'Z': // Timezone name
+            // return write_str(writer, to_conv, timeptr);
+
+    // composite conversions
+  case 'c': // locale specified date and time
+  case 'D': // %m/%d/%y (month/day/year)
+  case 'F': // %Y-%m-%d (year-month-day)
+  case 'r': // %I:%M:%S %p (hour:minute:second AM/PM)
+  case 'R': // %H:%M (hour:minute)
+  case 'T': // %H:%M:%S (hour:minute:second)
+  case 'x': // locale specified date
+  case 'X': // locale specified time
+            // return write_composite(writer, to_conv, timeptr);
+  default:
+    return writer->write(to_conv.raw_string);
+  }
+  return 0;
+}
+
+} // namespace strftime_core
+} // namespace LIBC_NAMESPACE_DECL