[k2] add Date/Time classes to light runtime #1481

Shamzik · 2025-12-08T16:26:17Z

This first part of implementing Date/Time classes. Here implemented:

DateInterval;
DateTimeZone;
DateTimeImmutable;
DateTime.

builtin-functions/kphp-light/stdlib/time-functions.txt

apolyakov · 2025-12-11T19:02:44Z

runtime-common/core/core-types/decl/string_decl.inl

 class string {
 public:
  using size_type = string_size_type;
+  using value_type = char;


What's the purpose of this change?

This was added to be able to use std::back_inserter

Let's not change core KPHP types

runtime-light/stdlib/time/time-functions.cpp

apolyakov

I've reviewed and noted some comments. Please, take into account that these comments should be applied to whole PR, not only the places where I left those comments

apolyakov · 2025-12-12T10:00:12Z

runtime-common/stdlib/string/string-context.h

 inline constexpr std::string_view BACKSLASH_ = "\\";
 inline constexpr std::string_view QUOTE_ = "\"";
 inline constexpr std::string_view NEWLINE_ = "\n";
+inline constexpr std::string_view NOW_ = "now";


It's the context for strings part of standard library. Let's move this to one specific for time functions

apolyakov · 2025-12-26T12:22:04Z

runtime-light/stdlib/time/timelib-functions.h


 #include "kphp/timelib/timelib.h"

+#include "common/containers/final_action.h"


Please, remove all the unused includes

apolyakov · 2025-12-26T12:22:56Z

runtime-light/stdlib/time/timelib-functions.h

+
+using error_container_t = std::unique_ptr<timelib_error_container, destructor>;
+using rel_time_t = std::unique_ptr<timelib_rel_time, destructor>;
+using time_t = std::unique_ptr<timelib_time, destructor>;


Now we have a mix of raw timelib types and our own wrappers in the API. It doesn't look like a good solution.

Also, please refactor exisiting code that uses raw types in case you add wrappers. Mixing types is always bad

fixed except clone function

apolyakov · 2025-12-26T12:23:30Z

runtime-light/stdlib/time/timelib-functions.h

+
+time_offset_t construct_time_offset(timelib_time& t) noexcept;
+
+std::pair<time_t, error_container_t> parse_time(std::string_view time_sv) noexcept;


Is it possible for us to have both valid time_t and error_container_t?

error_container_t is always non-null here. time_t is null in case of an error.

So, it seems that std::expected<std::pair<time, error>, error> is the right choice here

runtime-light/stdlib/time/timelib-functions.h

apolyakov · 2025-12-26T12:26:38Z

runtime-light/stdlib/time/timelib-functions.h

+
+time_t add(timelib_time& t, timelib_rel_time& interval) noexcept;
+
+rel_time_t clone(timelib_rel_time& rt) noexcept;


Isn't it better to just create a wrapper class? As far as I see, it might be worth it since you already have clone and destruct

I've refactored the header to make it easier to read. This library isn't very friendly to creating wrapper classes around it.

runtime-light/stdlib/time/timelib-functions.h

apolyakov · 2026-01-17T17:56:48Z

runtime-light/stdlib/time/timelib-constants.h


 } // namespace timezones

+inline constexpr std::string_view NOW_ = "now";


It doesn't look like it should be public

apolyakov · 2026-01-17T17:57:59Z

runtime-light/stdlib/time/timelib-functions.h

+constexpr inline std::array<std::string_view, 7> DAY_FULL_NAMES = {"Sunday", "Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday"};
+constexpr inline std::array<std::string_view, 7> DAY_SHORT_NAMES = {"Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"};
+
+struct error_container_destructor {


Is it supposed to be manually used by a user of kphp::timelib?

apolyakov · 2026-01-17T17:58:26Z

runtime-light/stdlib/time/timelib-functions.h

+  }
+};
+
+using error_container_t = std::unique_ptr<timelib_error_container, kphp::timelib::error_container_destructor>;


I'd suggest to get rid of _t suffix

apolyakov · 2026-01-17T18:00:26Z

runtime-light/stdlib/time/timelib-functions.h

+using time_t = std::unique_ptr<timelib_time, kphp::timelib::time_destructor>;
+
+/* === rel_time === */
+std::expected<kphp::timelib::rel_time_t, kphp::timelib::error_container_t> parse_interval(std::string_view format) noexcept;


I don't understand what does this function do. Its name is parse_interval while it accepts some format? What does format mean here?

apolyakov · 2026-01-17T18:02:11Z

runtime-light/stdlib/time/timelib-functions.h

+
+/* === rel_time === */
+std::expected<kphp::timelib::rel_time_t, kphp::timelib::error_container_t> parse_interval(std::string_view format) noexcept;
+kphp::timelib::rel_time_t clone_time_interval(const kphp::timelib::time_t& t) noexcept;


Is kphp::timelib::time_t an interval? It seems more like a timepoint

apolyakov · 2026-01-17T18:14:12Z

runtime-common/core/core-types/decl/string_decl.inl

 class string {
 public:
  using size_type = string_size_type;
+  using value_type = char;


Let's not change core KPHP types

apolyakov · 2026-01-17T18:14:42Z

runtime-light/stdlib/time/timelib-functions.h

+  timelib_sll isoweek{};
+  timelib_sll isoyear{};
+
+  for (std::size_t i{0}; i < format.size(); ++i) {


We don't use std:: for primitive types

apolyakov · 2026-01-17T18:16:12Z

runtime-light/stdlib/time/timelib-functions.h

+
+template<typename OutputIt>
+OutputIt format_to(OutputIt out, std::string_view format, const kphp::timelib::time_t& t) noexcept {
+  return kphp::timelib::details::format_to<OutputIt>(out, format, t, t->is_localtime);


Why the approach to implement format_to for time and rel_time differs?

apolyakov · 2026-01-17T18:17:26Z

runtime-light/stdlib/time/timelib-functions.h

+    options |= TIMELIB_OVERRIDE_TIME;
+  }
+
+  kphp::memory::libc_alloc_guard{}, timelib_fill_holes(time.get(), now_time.get(), options);


libc_alloc_guard{} is not free in terms of performance. Try to use it as less as possible

apolyakov · 2026-01-17T18:20:05Z

runtime-light/stdlib/time/timelib-functions.h

+    case 'o':
+      if (!weekYearSet) {
+        timelib_isoweek_from_date(t->y, t->m, t->d, std::addressof(isoweek), std::addressof(isoyear));
+        weekYearSet = 1;


Implicit cast here

apolyakov · 2026-01-17T20:08:56Z

runtime-light/stdlib/time/time-state.h

+  }
+
+  void update_last_errors(kphp::timelib::error_container_t&& new_errors) noexcept {
+    last_errors.swap(new_errors);


Is seems unnecessary to waste CPU cycles to move a value back to new_errors

apolyakov · 2026-01-17T20:14:09Z

runtime-light/stdlib/time/datetimezone.h

@@ -0,0 +1,34 @@
+// Compiler for PHP (aka KPHP)
+// Copyright (c) 2022 LLC «V Kontakte»


Please, do not let copy paste introduce errors. Change the year to 2025

apolyakov · 2026-01-17T20:15:04Z

runtime-light/stdlib/time/datetimezone.h

Let's follow our style and name files like datetime-zone.h

apolyakov · 2026-01-17T20:16:19Z

runtime-light/stdlib/time/datetimezone.cpp

+  return self;
+}
+
+string f$DateTimeZone$$getName(const class_instance<C$DateTimeZone>& self) noexcept {


What was the reason to move these definitions into the cpp file?

apolyakov · 2026-01-17T20:18:21Z

runtime-light/stdlib/time/dateinterval.cpp

+    kphp::log::warning("DateInterval::createFromDateString(): Unknown or bad format ({}) {}", datetime.c_str(), errors);
+    return {};
+  }
+  class_instance<C$DateInterval> date_interval;


There is make_instance alternative

Shamzik marked this pull request as ready for review December 10, 2025 10:26

Shamzik force-pushed the kshamazov/datetime branch from d1df9ec to 244c045 Compare December 10, 2025 22:05

Shamzik self-assigned this Dec 10, 2025

Shamzik added the k2 k2 related label Dec 10, 2025

Shamzik added this to the next milestone Dec 10, 2025

Shamzik changed the title ~~[k2] add Date/Time classes (except DateTime) to light runtime~~ [k2] add Date/Time classes to light runtime Dec 10, 2025

Shamzik requested review from PetrShumilov, apolyakov and astrophysik December 11, 2025 16:17

apolyakov reviewed Dec 11, 2025

View reviewed changes

Shamzik requested a review from apolyakov December 12, 2025 11:44

DrDet modified the milestones: 14.12.2025, next Dec 12, 2025

Shamzik added 17 commits December 16, 2025 12:18

stash

91759b3

stash

c35762c

stash

01f2476

stash

b29deda

fixes

6c3b902

datetime methods

d9da675

kphp::memory::libc_alloc_guard

9825b2b

some tests

7161d44

fixes

283b554

format

a7ac9db

get_last_errors

40b60bd

error messages

d5bb003

test 12_create_interval.php

6273913

remove extra ';'

cc608cb

auto -> int32_t

028e647

fixes

dc43023

implement setISODate method

5362352

Shamzik added 2 commits December 16, 2025 12:18

implement sub methods

9b7918d

implement getOffset methods

a96050c

Shamzik force-pushed the kshamazov/datetime branch from 49fc0f4 to a96050c Compare December 16, 2025 09:20

Shamzik added 2 commits December 25, 2025 16:16

construct -> parse

39cfffc

use destructor{}

befbf8d

apolyakov requested changes Dec 26, 2025

View reviewed changes

Shamzik added 5 commits January 14, 2026 18:14

move NOW_STR into time state

b55795f

remove include

9cc1a77

use type wrappers

22fede1

fix parse_interval

c2e2dd4

refactoring

18a490d

Shamzik requested a review from apolyakov January 16, 2026 16:14

destructors

589fd73

apolyakov requested changes Jan 17, 2026

View reviewed changes

apolyakov reviewed Jan 17, 2026

View reviewed changes


		#include "kphp/timelib/timelib.h"

		#include "common/containers/final_action.h"


		time_offset_t construct_time_offset(timelib_time& t) noexcept;

		std::pair<time_t, error_container_t> parse_time(std::string_view time_sv) noexcept;


		time_t add(timelib_time& t, timelib_rel_time& interval) noexcept;

		rel_time_t clone(timelib_rel_time& rt) noexcept;


		} // namespace timezones

		inline constexpr std::string_view NOW_ = "now";

		@@ -0,0 +1,34 @@
		// Compiler for PHP (aka KPHP)
		// Copyright (c) 2022 LLC «V Kontakte»

[k2] add Date/Time classes to light runtime #1481

Are you sure you want to change the base?

[k2] add Date/Time classes to light runtime #1481

Conversation

Shamzik commented Dec 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

apolyakov left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Shamzik commented Dec 8, 2025 •

edited

Loading