Mihrab Methods

Verification

Every method in Mihrab Methods is compared, time by time, with the timetable its authority publishes. This page says precisely what that means, so you can decide how far to rely on it.

In one line: 77,661,385 printed times compared across 21 methods; 0 differ. Last checked 17 November 2026.

What is compared

One time is one prayer time on one day in one place: Fajr, sunrise, Dhuhr, Asr, Maghrib or Isha, as a clock time to the minute. For each authority, the library computes every place the authority lists on every day it has published, and each computed time is compared with the minute printed in the authority’s timetable, on the clock that timetable is printed in.

Tolerance: none

A time matches only if it is the same minute. There is no “within one minute” allowance: a time one minute away from the printed one is counted as off. A method is included in the library only when its count of times off is zero; a method at 99.9 % is not shipped.

Results by authority

Authority Tables matched Places Times checked Off the printed minute By the method alone Correction entries Checked over
Sweden Islamiska Förbundet’s tables 107 65,442 0 99.969 % 270 11 Oct 2026 → 21 Dec 2027
  Diyanet’s own tables 159 1,393,794 0 99.981 %   1 Jan 2024 → 31 Dec 2027
Diyanet its tables worldwide 8,318 71,630,078 0 99.987 % 9,101 1 Jan 2024 → 31 Dec 2027
Morocco Habous 164 94,350 0 99.885 % 108 27 Aug 2026 → 30 Nov 2026
Algeria Ministry of Religious Affairs 68 101,184 0 99.365 % 642 1 Oct 2026 → 5 Jun 2027
Indonesia Kemenag 517 362,934 0 99.935 % 234 7 Oct 2026 → 31 Jan 2027
Türkiye Takvimi its cities’ years 821 1,812,768 0 99.952 % 860 31 Dec 2025 → 2 Jan 2027
Kazakhstan the Spiritual Administration’s times 84 551,880 0 100.000 % 0 1 Jan 2025 → 31 Dec 2027
Egypt the Survey Authority’s times 77 168,630 0 99.373 % 1,056 1 Jan 2026 → 31 Dec 2026
Brunei the Ministry’s times 4 17,064 0 99.671 % 56 9 Jan 2025 → 31 Dec 2026
Qatar the Calendar House’s times 1 1,194 0 95.393 % 55 16 Jun 2026 → 31 Dec 2026
Kuwait the Ministry’s table 46 201,480 0 95.205 % 9,660 1 Jan 2026 → 31 Dec 2027
Jordan the Ministry’s times 17 37,230 0 98.380 % 603 1 Jan 2026 → 31 Dec 2026
Tatarstan the Administration’s timetable 44 96,360 0 98.071 % 1,858 1 Jan 2026 → 31 Dec 2026
Russia (Central Muftiate) the Central Muftiate’s timetables 18 39,342 0 99.877 % 48 1 Jan 2026 → 31 Dec 2026
Albania the Community’s calendar 22 48,180 0 98.082 % 924 1 Jan 2026 → 31 Dec 2026
Bulgaria the Grand Mufti’s Office’s timetable 48 105,120 0 97.671 % 2,448 1 Jan 2026 → 31 Dec 2026
Pakistan (Banuri Town) Jamia Banuri Town’s timetable 100 219,000 0 99.996 % 8 1 Jan 2026 → 31 Dec 2026
United Arab Emirates the Authority’s times 41 179,580 0 98.221 % 3,036 1 Jan 2026 → 31 Dec 2027
Iran (Tehran Institute of Geophysics) the Institute of Geophysics’ sheets 31 90,520 0 99.002 % 903 21 Mar 2025 → 20 Mar 2027
United Kingdom (Wifaqul Ulama) Wifaqul Ulama’s timetables 81 350,400 0 98.625 % 4,641 1 Jan 2026 → 31 Dec 2027
Tunisia the Institute’s times 26 94,855 0 97.727 % 2,156 1 Jan 2025 → 31 Dec 2026

21 of 21 methods certified; 0 of 77,661,385 times off the printed minute. Last checked 17 November 2026. Generated from certification.json by scripts/readme.mjs.

How to read the columns:

  • Places and Checked over are the coverage: how many of the authority’s listed places, and which dates. Coverage differs a lot between authorities, because some publish years ahead for thousands of places and some publish a few months for a few dozen. A short span is a smaller body of evidence, and the table shows it rather than hiding it.
  • Times checked is the number of individual comparisons.
  • Off the printed minute is the result with the correction list applied. It is zero for every method, by the rule above.
  • By the method alone is the share that matches from the formula with nothing added.
  • Correction entries is the size of the list that covers the rest.

The correction lists — the part to understand

No formula reproduces a printed timetable completely, because an authority’s own program rounds numbers that sit a fraction of a second from a minute boundary, and which side they fall on depends on details of its arithmetic that cannot be seen from outside. The method here is never more than a minute from such a time, and each one is recorded in a per-authority correction list: place, date, prayer, one minute up or down.

So “zero off” is the formula plus that list. The “By the method alone” column is the honest measure of the formula itself, and the “Correction entries” column is the size of the remainder. A list is kept small by rule: if it grows beyond a small fraction of the times, the formula is treated as unfinished. The lists are in data/corrections, ship inside the package, and can be refreshed by an app without a release (the feed).

Known exceptions and limits

  • Only the dates compared are verified. Inside them an answer carries status: 'certified'; outside them 'method-only' — the same formula, no correction list, not compared with anything. certifiedFrom and certifiedUntil are on every answer. Past dates before the span are in the same position as future ones.
  • Only the listed places. A method is verified at the places its authority publishes for. The library answers for those places and tells you how far the nearest one is; it does not claim a verified time for an arbitrary coordinate.
  • Rows set aside. Where a timetable’s own row is evidently faulty — another town’s times under a name, a row repeated from another day or year — the row is set aside rather than counted, and the number set aside is recorded per check in the record. They are listed per authority in where each method stands.
  • The Asr a timetable does not print. Each timetable prints one Asr. The other school’s Asr is available on request but was compared with nothing; its basis is 'unchecked'.
  • Standard conventions are not verified. standardTimes (MWL, ISNA, Egypt, Karachi, Umm al-Qura) is the plain astronomical calculation for places no authority here covers, and is marked 'calculated'.
  • Authorities change their timetables. A republished table can differ from the one compared. The comparison is rerun weekly and whenever the tables change; a method that no longer matches loses its certification until it does.
  • Time zone rules. Clock times depend on the time-zone database of the device. Two devices with different editions can differ where a country’s rules changed between them.
  • Ports. The Python, Swift, Kotlin and Dart ports are held to a set of test vectors generated from the TypeScript library on every push, not each separately to the full timetables.
  • Not every authority. Authorities whose methods have not reached zero off are absent. Status lists what is missing.

The record

The result of each run is written to certification.json: per authority and per timetable, the places, the times, how many are off with and without the correction list, the dates, the rows set aside, the method’s revision and the SHA-256 of its correction list. Every figure on this site and in the README is generated from that file; none is typed by hand. The same record is served at /v1/methods by the API and by methodInfo() in the library.

The comparison is tools/verify.ts; the record is written by tools/certify.ts.

What you can check yourself

The authorities’ timetables are their own publications and are not redistributed in the repository, so the full comparison needs the timetables supplied to it (the check reads them from the directory named by MIHRAB_TABLES). Without them you can still:

  • run npm test, which holds the library to test/vectors.json and checks the record against the correction lists’ hashes;
  • compare any day with the timetable your mosque follows — and report a difference with the city, date, prayer and printed time;
  • read each method, formula by formula, in the methods reference.

Adding an authority

The rule is “100 % or nothing”. A new authority needs:

  1. The formula, written from scratch from the authority’s published times: its sun, angles, margins, rounding and high-latitude rule. Nothing is copied from an authority’s software.
  2. The places, each at the point its table is computed at, with the clock the table is printed in.
  3. The check against the authority’s published tables: every place, every day.
  4. A correction list only for what rounding leaves.
  5. Certification at zero off, a written description of the method in the methods reference, and test vectors so every port can be held to it.

The steps and commands are in CONTRIBUTING.md. If your country’s authority is missing, an issue naming it is a good start.

Mihrab Methods is independent of every authority named here: it matches their published timetables and speaks for none of them.

Edit this page on GitHub · Report a problem