| Work / author / dates | al-Risāla al-Qushayriyya (الرسالة القشيرية), "The Epistle on Sufism" · Abū l-Qāsim ʿAbd al-Karīm b. Hawāzin al-Qushayrī (376–465 AH / 986–1072 CE), written 437/1045 |
| Part added | A first selection, in the book's order: (1) the chapter on the terms («باب في تفسير ألفاظ تدور بين هذه الطائفة وبيان ما يشكل منها»): the introduction and all 27 terms, from al-waqt to al-sirr; (2) the chapter on insight («باب الفراسة»); (3) the chapter on knowledge of God («باب المعرفة بالله»). Not included yet: the chapter on love («باب المحبة», OpenITI ll. 7963–8230, about 2,950 words), left for a later addition to keep this first text near 13,000 words; the creed, the 83 biographies and the other stations. |
| Id / files | qushayri · data/sources/qushayri-base.json (built by scripts/recitals/qushayri_base.py), qushayri-meta.json, qushayri.json (vowelled), qushayri-analysis.json (word ids qs:<passage>:<token>), AI layer app/src/content/ai/qushayri.json |
| Reader route | /read/qushayri · about /read/qushayri/about |
| Status | working edition from one digital text, established 2 Oct 2026 by Claude · reviewed by a person: — |
1. Witnesses
| Siglum | Witness | Kind | Vowelled? | Licence | Location |
|---|---|---|---|---|---|
| O | OpenITI 0465IbnHawazinQushayri.RisalaQushayriyya.Shamela0009953-ara1, from al-Maktaba al-Shāmila (book 9953). The print behind it, as OpenITI's metadata records it: ed. ʿAbd al-Ḥalīm Maḥmūd & Maḥmūd b. al-Sharīf, Cairo: Dār al-Maʿārif, 2 vols (year and printing not recorded). 113,891 tokens. Page tags give vol:page of that edition. | digital (of a modern edition) | no | OpenITI corpus CC BY-NC-SA 4.0; the edition is © as an edition | GitHub OpenITI/0475AH, pinned at commit cc9beb0d9645e8cdce892556423b9926a127054f: the file |
| J | OpenITI's other version, …RisalaQushayriyya.JK009321-ara1 (ed. Khalīl al-Manṣūr, Beirut: Dār al-Kutub al-ʿIlmiyya, 1422/2001, per its metadata; OpenITI marks it PRIMARY_VERSION). 113,782 tokens. | digital (of a modern edition) | no | CC BY-NC-SA 4.0 | same repository and commit: the file. Collated for the selection (§3); nothing adopted |
| — | Translations, all in copyright, not used for our English: Knysh, Al-Qushayri's Epistle on Sufism (Garnet 2007); von Schlegell, Principles of Sufism (1990, the stations); Sells, Early Islamic Mysticism (1996, excerpts). R. Hartmann, Al-Kuschairis Darstellung des Ṣûfîtums (Berlin 1914, German, partial) is public domain but no scan was located. | © (Hartmann PD) | Knysh may be compared privately, by locator only |
2. Base and why
O, the Maḥmūd–Sharīf edition: the edition scholars cite, and the version docs/research/heart-sight.md §3.2 already quotes (so its line links stay valid). There is no public-domain print in hand; a PD witness (e.g. the Būlāq 1284/1867 print, or the Cairo prints with Zakariyyāʾ al-Anṣārī's commentary) would be the next step if a person wants the text settled.
Selection (OpenITI lines, page tags close a page, so a paragraph's page is the first tag after it):
| Part | OpenITI lines | vol:page | Passages |
|---|---|---|---|
| The terms: introduction + 27 terms | 1760–2523 | I:150–206 | 0–77 |
| Insight (al-firāsa) | 5882–6124 | II:386–396 | 78–104 |
| Knowledge of God (al-maʿrifa bi-llāh) | 7812–7962 | II:477–483 | 105–119 |
The 27 terms (each starts a passage and a heading in the reader): al-waqt; al-maqām; al-ḥāl; al-qabḍ wa-l-basṭ; al-hayba wa-l-uns; al-tawājud, al-wajd, al-wujūd; al-jamʿ wa-l-farq; al-fanāʾ wa-l-baqāʾ; al-ghayba wa-l-ḥuḍūr; al-ṣaḥw wa-l-sukr; al-dhawq wa-l-shurb; al-maḥw wa-l-ithbāt; al-satr wa-l-tajallī; al-muḥāḍara, al-mukāshafa, al-mushāhada; al-lawāʾiḥ, al-ṭawāliʿ, al-lawāmiʿ; al-bawādih wa-l-hujūm; al-talwīn wa-l-tamkīn; al-qurb wa-l-buʿd; al-sharīʿa wa-l-ḥaqīqa; al-nafas; al-khawāṭir; ʿilm/ʿayn/ḥaqq al-yaqīn; al-wārid; al-shāhid; al-nafs; al-rūḥ; al-sirr. There is no chapter on qalb: the heart is treated under nafs/rūḥ/sirr, khawāṭir and firāsa.
3. Collation
O against J, for the whole selection (2 October 2026; Claude, by script, then the larger differences read in context). No print and no manuscript compared: both witnesses are digital texts of modern editions, so this is a check of O's transmission and typesetting, not a critical apparatus of the work. No reading of O was changed (§4); every difference is recorded.
- Method: both files at the pinned commit; the selection in O (lines 1760–2523, 5882–6124, 7812–7962) aligned with J's chapters (J lines 1575–2285, 5315–5545, 7079–7206); OpenITI tags, the edition's braces and brackets, punctuation and any vowel signs removed; aligned word by word (Python
difflib). Script: kept in the session scratchpad only (to be added toscripts/recitals/if the comparison is rerun). - Full list:
docs/texts/qushayri-collation-JK009321.tsv(804 differences, each with our passage number, the line in O and in J, a category, and both readings). Categories: orthography only (hamza seats, alif maqṣūra / yāʾ, tāʾ marbūṭa) 240; one letter 308; word variant 135; one- or two-word plus/minus 102; longer omissions or additions (three words or more) 19. - On the whole J is the more careless copy (dropped letters and words: «أ واله» for أحواله, «فنما» for فإنما, «الشيكان»; two dittographies, at ¶3 and ¶61/¶65; omissions at ¶0, ¶20, ¶52, ¶81, ¶111). But O has its own faults, and at a number of places J has the reading the sense requires. Those places, read in context, are listed below as apparatus: O's reading (what we print), J's reading, and an assessment. The assessment is ours (AI), not an editor's; nothing is adopted until a person decides, ideally with a printed witness (the Būlāq 1284 print, or Maḥmūd–Sharīf's printed volumes, to see whether O's faults are the edition's or the digitiser's).
Apparatus: places where O is probably faulty (O = our base)
| ¶ | O line / J line | O (printed) | J | Assessment |
|---|---|---|---|---|
| 4 | 1817 / 1626 | رؤية التفسير فيها | رؤية التقصير فيها | J: "seeing [one's] shortfall in them"; O's التفسير makes no sense here. Probable O slip |
| 4 | 1818 / 1628 | صبيانهم عن محل الإعجاب | صيانتهم عن محل الإعجاب | J: "to guard them from self-admiration". O's "their boys" is a slip; our AI English already reads J ("to protect them") |
| 5 | 1823 / 1634 | تأتي من غير الوجود | تأتي من عين الجواد | Neither is clean: J "from the source of the Bountiful" makes the contrast with "effort" (المجهود) and rhymes; O "from other than wujūd" is probably a corruption of a form like عين الجود. Our AI English renders O ("apart from any finding of one's own") |
| 9 | 1863–1865 / 1668–1669 | the clause «يوجب قبضا … لا مساغ لغير وارده» twice | once | Dittography in O: our ¶9 prints the clause twice. Probable O error |
| 12 | 1889 / 1691 | مطلق القبض والبسط | حالتي القبض والبسط | Both possible ("contraction and expansion as such" / "the two states of …"); undecided |
| 31 | 2071 / 1856–1857 | فإذا فني عن شهواته بقي بصدق إنابته، ومن عالج أخلاقه | فإذا فنى عن شهواته بقي بنيته وإخلاصه في عبوديته. ومن زهد في دنياه بقلبه يقال فنى عن رغبته فإذا فني عن رغبته فيها بقي … | J has a whole further step (renouncing the world in the heart) that fits the sequence of fanāʾ stages; O probably lost it by eye-skip (فإذا فني عن … / فإذا فني عن …). The wording of the remainder (بصدق إنابته / بنيته وإخلاصه) differs too. Undecided without a print |
| 35 | 2116–2120 / 1901–1904 | الشلبي (four times) | الشبلي | O's typo for al-Shiblī; our AI English already has al-Shiblī |
| 39 | 2144 / 1925 | فقال: أريد أبا يزيد؟ وأين أبا يزيد | فقال: أريد أبا يزيد. فقال: من أبو يزيد؟ وأين أبو يزيد؟ | J: the visitor asks for Abū Yazīd, and Abū Yazīd answers "Who is Abū Yazīd?". O lost the answer's opening (eye-skip on أبا يزيد / أبو يزيد) and has the wrong case after أين. Probable O omission |
| 42 | 2182 / 1958 | فصاحب الذوق متساكر وصاحب الري صاح | فصاحب الذوق متساكر، وصاحب الشرب سكران، وصاحب الري صاح | The passage sets out three degrees, dhawq, shurb and riyy (the sentence before names all three); J's three members match it. Probable O omission |
| 58 | 2337 / 2098–2100 | مكنه الحق سبحانه في كل نفس | مكنه الحق سبحانه، بأن لا يرده إلى معلولات النفس، فهو متمكن في حاله، على حسب محله واستحقاقه. ثم يتحفه الحق سبحانه في كل نفس | Eye-skip in O on الحق سبحانه … الحق سبحانه. Probable O omission |
| 62 | 2389 / 2149 | جلت العملية عن قبول الوصل والفصل | جلت صمديته عن قبول الوصل والفصل | J: "His eternal self-sufficiency is too majestic to admit joining or separation"; O's "the practical" is a slip. Our AI English already reads J ("[His eternal Self-Sufficiency]", bracketed) |
| 65 | 2415 / 2175 | فقد يكون بإلقاء الشيطان | وهو قد يكون بإلقاء ملك، وقد يكون بإلقاء شيطان | O goes on "فإذا كان من الملك …", so it needs the angel here. Probable O omission (eye-skip on بإلقاء) |
| 81 | 5920 / 5350 | والذين هم أكثر من حظا | والذين هم أكبر منه حظا | O's «من حظا» is ungrammatical; J (or أكثر منه) wanted. O slip |
| 99 | 6080 / 5503 | حدثني أخي أنها كانت حاملا بي | حدثتني أمي أنها كانت حاملا بي | "My mother told me that she was pregnant with me": O's "my brother … she" is impossible. Our AI English papers over O ("My brother told me that when she [our mother] was pregnant…"); it should follow J, with a note. Probable O error |
Apparatus: places where J is faulty (recorded so they are not adopted by mistake): ¶0 (J omits أو مجلوبة بضرب تصرف بل هي معان أودعها الله تعالى قلوب قوم, leaving the next clause without its parallel); ¶3 and ¶61, ¶65 (dittographies in J); ¶20 (J omits التواجد يوجب استيعاب العبد والوجد, so it has two of the three degrees); ¶52 (J omits فاللوائح كالبروق ما ظهرت حتى استترت); ¶63 (J garbled: «غير مؤيدة مؤيدة … فالشريعة جاءت غير مقيدة»); ¶81 (J omits the verse Q 3:79 and its gloss); ¶111 (J omits Sahl's saying on bewilderment).
Undecidable on these two witnesses: proper names, ¶97 عبيد / سعيد and ¶110 عباس / عياش; ¶80 «إلا أن يكون لنفسه بها حظ فتحجب عن الله تعالى» / «… لنفسك فيها حظ فتحجب عن حاجتك»; ¶119, the order of the last two sayings of the maʿrifa chapter (J has Muḥammad b. al-Faḍl's "knowledge is the life of the heart with God" before the question to al-Kharrāz, O after it).
4. Decisions
No reading of O changed, including the probable faults of O found by the collation (§3): they are recorded there, pending a person's decision. Note that the AI English already departs from O, unrecorded until now, at ¶4 (صيانتهم), ¶35 (al-Shiblī) and ¶62 (صمديته), where it silently follows the sense J has; and at ¶99 it renders O's impossible «حدثني أخي أنها» with an inserted "[our mother]". Mechanical (recorded in qushayri-base.json → mechanical and changes):
- OpenITI page tags and milestones (
msNNN) removed. - The edition's apparatus not reproduced: the curly braces round Quran quotations and the bracketed references «[البقرة: 245]» removed (28, each logged with its OpenITI line).
- Shamela's « , » written as the Arabic comma «،».
- Very long paragraphs (over 170 words) cut into two passages at the sentence end nearest the middle; short paragraphs joined to the passage before.
5. Quotations verified
28 Quran quotations, all marked by the edition, replaced by the verbatim Tanzil Uthmani text (scripts/recitals/check_quotes.py qushayri: 28 verified, 0 failures). One spelling: Q 2:245 «وَيَبْصُۜطُ» (Tanzil, with ṣād read as sīn) for the edition's يبسط. Quran allusions not marked by the edition were not searched for by hand; scripts/quran_refs.py (skeleton matcher) runs over the text.
Hadith, with their grading shown, not implied (quotes, kind hadith):
- maʿrifa, passage 105: «إن دعامة البيت أساسه، ودعامة الدين المعرفة بالله تعالى واليقين والعقل القامع», with the isnād through ʿAbbād b. Kathīr, judged very weak (matrūk) by the hadith critics [REPORTED]: a weak hadith, not to be cited as a sound Prophetic saying.
- firāsa, passage 78: «اتقوا فراسة المؤمن فإنه ينظر بنور الله», al-Tirmidhī, Tafsīr, sūrat al-Ḥijr (no. 3127), who calls it gharīb; its chain (ʿAṭiyya al-ʿAwfī) is weak [REPORTED].
- In the reader, each grading is also on the hadith's first word as a term note ("Hadith grading", in
app/src/content/ai/qushayri.json, AI-badged; added aftertranslate.py, so a rerun must keep it). - The many sayings of the masters (al-Junayd, al-Shiblī, al-Wāsiṭī, Abū ʿAlī al-Daqqāq, …) are the book's own reports, with their isnāds as printed; not checked.
6. Vowelling
O is unvowelled. Full vowelling by AI (Claude Opus, scripts/recitals/prepare.py qushayri, mask a), letters verified unchanged; the Quran quotations restored verbatim from Tanzil afterwards (restore_quran.py, mask t). Ambiguous words are listed per passage in qushayri.json → ambiguous. No vowelled edition compared.
Review of the names and of the vowel conflicts (2 October 2026; Claude, not yet a person). Scope: every personal name, kunya, nasab chain and nisba in the selection (about 190 name phrases, the isnāds of ¶13, ¶23, ¶78, ¶99, ¶105, ¶107, ¶110–112, ¶115–119 included), and all 523 "vowel conflicts" (words whose AI vowelling differs from CAMeL's best analysis). References: al-Samʿānī, al-Ansāb (OpenITI 0562Samcani.Ansab.Shamela0012317-ara1, which spells each nisba out in words: "bi-ḍamm al-ṣād …"), al-Dhahabī, Siyar aʿlām al-nubalāʾ (OpenITI 0748Dhahabi.SiyarAclamNubala.Shamela0010906-ara1, unvowelled, for the identity and spelling of the men), and Arabic grammar (case agreement of ibn and of the kunya in every chain was checked by script: no error). The vowels only were reviewed: the letters are O's and stay O's (§4), so a name misspelt in O keeps its misspelling, vowelled as written.
- The 523 conflicts. Nearly all are CAMeL's limits, not errors of ours: names and nisbas it does not know (it leaves them bare, about 150), passives it reads as actives (ḥukiya, rūwiya, ubtuliya, ḥujibtu, yuḥmalu, sutira, ruddū …), satr "veiling" read as sitr, ḥaddathanā without its shadda, shirb "a drinking share" read as shurb, poetry it cannot scan (sukrāni "two intoxications", ʿuqār, mudhhab), and the case of ibn. One clear error of ours was found among them, and one inconsistency (below).
- Checked against al-Samʿānī and found right: al-Ṣuʿlūkī (bi-ḍamm al-ṣād … wa-ḍamm al-lām), al-Tīnātī (bi-kasr al-tāʾ), al-Wajīhī, al-Shajarī, al-Sūsī, al-Murtaʿish, al-Kattānī, al-Naṣrābādhī (O writes النصرأباذي, so the AI's hamza on the second alif is O's letter), al-Kirmānī (al-Samʿānī: properly Karmān, but "known with kasra", as we print), and al-Qirmiṭī in ¶96 (al-Samʿānī vowels the nisba of the Qarāmiṭa bi-kasr al-qāf … wa-kasr al-mīm, as our AI did; al-Qarmaṭī is the commoner modern vowelling). The other names are the usual vowellings of well-known men (al-Junayd, al-Shiblī, al-Jurayrī, Ibn Masrūq, al-Nūrī, Abū Ḥamza, al-Ḥīrī, al-Duqqī, al-Daylamī, al-Būshanjī, al-Khuldī, al-Rūdhbārī, al-Dārānī, al-Nahrajūrī, al-Sulamī, Ibn Nujayd, Ibn Yazdānyār, al-Anṭākī, Ibn Abī l-Ḥawārī, al-Wāsiṭī, Shāh al-Kirmānī …) and were left as they are.
Changes (in data/sources/qushayri.json; mask kept a; analysis rebuilt with scripts/fusus/analyze.py):
| ¶ | word id | was | now | why |
|---|---|---|---|---|
| 40 | qs:40:134 | فَضَحْوُكَ | فَضَحُوكَ | faḍaḥūka "they exposed you": the AI put sukūn on the ḥāʾ and ḍamma on the wāw of the plural ending |
| 42 | qs:42:145 | يَحْيي | يَحْيَي | Yaḥyā b. Muʿādh: O writes the final alif maqṣūra as yāʾ; the fatḥa before it was missing here (¶117 already had it) |
| 113 | qs:113:69 | يَحْيي | يَحْيَي | the same |
Doubtful, left for a person (all but the last are O's letters; the vowels follow them):
- ¶23, ¶90 «التُّرُوغَنْدِيّ» (O; J has النروغندي): al-Samʿānī has al-Turūghbadhī (التروغبذي, bi-ḍamm al-tāʾ wa-l-rāʾ … wa-fatḥ al-bāʾ), of Turūghbadh, a village of Ṭūs, which fits ¶90 (he sets out for Ṭūs). O's ـغندي is probably a misreading of ـغبذي.
- ¶10 «القَحْطِيّ» (O and J): probably al-Qaḥṭabī (al-Samʿānī knows the nisba القحطبي, bi-fatḥ al-qāf … wa-fatḥ al-ṭāʾ); not settled.
- ¶34 «الرَّبِيعَ بْنَ خَيْثَمٍ» (O and J): the Successor is usually spelt Khuthaym (خُثَيْم); O's letters give Khaytham.
- ¶95 «أَبَا عُمَرَ بْنَ عُلْوَانَ»: ʿUlwān or ʿAlwān; the references to hand are unvowelled.
- ¶89 «الزُّبَيْدِيّ»: al-Zubaydī (the tribe) or al-Zabīdī (Zabīd in the Yemen, al-Samʿānī's entry); the man is not identified.
- ¶105 «سُلَيْمَانُ بْنُ عِيسَى الشَّجَرِيُّ»: vowelled as al-Samʿānī's nisba of al-Shajara near Medina; the man is not identified.
- ¶86 «أَبُو الْحَسَنِ النُّورِيُّ» beside «أَبُو الْحُسَيْنِ النُّورِيُّ» elsewhere: al-Nūrī is Abū l-Ḥusayn; O's letters at ¶86.
- ¶38 «أَبِي يَرِيدَ» (O's typo for أبي يزيد, J has the zāy); ¶35 «الشَّلْبِيُّ» (O's typo for al-Shiblī, §3); ¶63 «الرُّبُوِيَّةِ» (O, for al-rubūbiyya); ¶23 «غَبْيَتِهِ» (O, for ghaybatihi); ¶35 «سُجُوهِ» (O, for sujūdihi); ¶47 «فَأَضَافَةُ» (O, for fa-aḍāfahu); ¶60 «وَجَدْجتُكَ» (O; J وجدتك); ¶80 «أُقْعُدْ» (O writes the hamza of the imperative); ¶34 «الْمُحْمَاهَ» (O's hāʾ for tāʾ marbūṭa).
- Vowels open to a second reading: ¶6 «فَيُرَي بِهِ»; ¶23 «مِقْدَارَ مَنُونٍ» (two mann?); ¶49 «مَلْقِيٌّ بِذَاتِهِ»; ¶53 «مُوقَفَةٌ» (or muwaqqafa); ¶55 «مَرْحَلٍ»; ¶60 «قَدْ سَمَانِي»; ¶90 «خَرْ» (a place?); ¶91 «وَقْتَ هَاجِرٍ»; ¶21 «أَدَامَ أَخْدَمَ اللَّهُ».
7. Segmentation and word IDs
120 passages, 12,859 words (whitespace tokens of the base), our own division: whole OpenITI paragraphs grouped to about 80–150 words, a new passage at every term and chapter. Each passage in the base records chapter, openitiLines, openitiUrl (GitHub line anchors at the pinned commit) and page (vol:page). Headings (sections): 30 (the terms' introduction, the 27 terms, the two chapters). Word ids qs:<passage>:<token>.
8. Analysis layer
CAMeL Tools, scripts/fusus/analyze.py qushayri.json qushayri-analysis.json qs.
9. Commentaries
None. Classical commentaries exist (Zakariyyāʾ al-Anṣārī, Natāʾij al-afkār al-qudsiyya, with al-ʿArūsī's gloss); not on disk.
10. Translation
AI (Claude Opus, scripts/recitals/translate.py qushayri opus), marked AI everywhere, with a glossary of the terms in qushayri-meta.json. There is no public-domain English. The model is told not to reproduce or imitate Knysh, von Schlegell, Sells or Hartmann; none of them was given to it.
11. Rights
O: OpenITI, CC BY-NC-SA 4.0 (attribution: OpenITI, and the Maḥmūd–Sharīf edition via al-Maktaba al-Shāmila). The edition's apparatus is not reproduced. Knysh, von Schlegell, Sells: ©, private comparison by locator only, nothing in the repo.
12. Open questions / to review
- ~~Compare with OpenITI
JK009321(al-Manṣūr)~~: done 2 Oct 2026 (§3, and the TSV). To decide: the 14 places in the first apparatus table (12 probable faults of O, 2 undecided) (adopt J's reading with a note, or keep O and note J); fix the ¶9 dittography; correct the ¶99 English. Then a public-domain print, to settle which faults are the edition's and which the digitiser's. - ~~A person should read the AI vowelling of the isnāds and proper names~~: reviewed by Claude against al-Samʿānī and the Siyar, 2 Oct 2026 (§6: three fixes, a list of doubtful names). Still for a person: that list, and the verse lines.
- Add the chapter on love (maḥabba), then shawq; the "term ring" proposed in
docs/research/new-texts-scope.mdA.1 (each term with al-Jurjānī's Taʿrīfāt and al-Qāshānī's Iṣṭilāḥāt). - The other hadith in the selection are not graded yet.
- ~~Themes and resonances (protocol 16) not run~~: the script's stop on a missing file is fixed; the whole-text map was made on 2 Oct 2026 (
app/src/content/resonances/qushayri.json, shown on the About page). No passage layer: the Arabic reader for this text has no resonances switch (onlyFususReaderand the Greek reader do). heart-sight.md§3.2 gives the maʿrifa definition as II:476–77; by OpenITI's page tags (which close a page) it is on II:477.
13. Change log
- 2026-10-02: vowelling review of the names, kunyas, nisbas and isnāds and of the 523 vowel conflicts (§6): 3 vowel fixes (¶40, ¶42, ¶113), the doubtful ones listed; analysis rebuilt.
- 2026-10-02: collated against OpenITI JK009321 (al-Manṣūr) for the whole selection: 804 differences (
qushayri-collation-JK009321.tsv); apparatus of the significant ones in §3; nothing adopted. Themes and resonances (whole text) generated. - 2026-10-02: record created; base built from O (terms chapter, firāsa, maʿrifa), vowelled, analysed, AI-translated; catalogue kind "Sufi handbooks" added.