Repository navigation
stdlib: random spells random() as next_double and lacks the Random class, getstate/setstate, randbytes, triangular and the nine *variate functions — 11 of CPython's 26 names #2321
Description
Class
Contract (CLAUDE.md › Core & Stdlib Conventions): a stdlib module wraps .NET internally and exposes the Python API — list.append(), not Add(). random (src/Sharpy.Stdlib/spy/random_module.spy, spy-sourced; generated Random/Random.cs; docs docs/stdlib/random.md) exposes 11 names against CPython 3.12's 26 (random.__all__), and the one every Python program calls first is spelled differently.
Cells (measured @ f8c37f9: grep '^def ' on the .spy; python3 -c "import random; print(sorted(random.__all__))")
| # | CPython name | Sharpy | lane |
|---|---|---|---|
| 1 | random() |
spelled next_double() — random.random() is SPY0203 |
false-refusal · program |
| 2 | Random class (independent generator instances; random.Random(7).random()), SystemRandom |
absent — the module is functions over one process-wide generator (P54 #2250 made that generator locked and process-wide, as CPython's module-level one is) | false-refusal · program |
| 3 | getstate() / setstate() |
absent | false-refusal · program |
| 4 | randbytes(n), triangular(low, high, mode) |
absent | false-refusal · program |
| 5 | betavariate, binomialvariate, expovariate, gammavariate, lognormvariate, normalvariate, paretovariate, vonmisesvariate, weibullvariate |
absent (gauss is present) |
false-refusal · program |
Present and named as CPython names them: seed, randint, uniform, choice, shuffle, randrange (2 overloads), gauss, getrandbits, choices, sample.
Seeded sequences differ from CPython's (System.Random vs Mersenne Twister) — that is a documented deviation, not a cell here; docs/design/python-fidelity-scope.md compares the API SURFACE, and a Python programmer's random.random() is the ordinary program that hits cell 1.
Acceptance
- Cell 1:
random.random()exists;next_doubleis removed or kept only as the documented .NET-flavoured alias (one way to do it — CLAUDE.md anti-patterns say one, so removal, with the generated docs regenerated). - Cells 2–5: either implemented at rung 1 (the
.spymodule — aRandomclass holding its own lockedSystem.Random, the variates overgauss/uniform) or recorded as adocs/deviations.yamlrow per name with the reason; no silent absence. - The regenerated
Random.cs,docs/stdlib/random.mdand the spy tests (Stdlib.Tests/Spy/random/…) in the same commit;StdlibModuleGlobalRosterTestsstill rosters exactly the written globals (aRandomclass instance field is per-instance, not a module global).
Found by the P54 /verify-implementation regression auditor (probe (c) in its brief: import random; random.seed(7); print(random.random()) vs python3) @ 045ce6b; filed on the owner's instruction 2026-10-15.
Activity
antonsynd commentedon Oct 15, 2026
Batching 2026-10-15 @ f8c37f9 — charter
.claude/plans/remediation-round-2026-09-20-batching.md§3.0 (cell 1) and §3.12d (cells 2–5), chore now + P61 — canonical stdlib & the ideal sweep (order 12d). Ruling R-HL (owner, 2026-10-15): (a) cell 1 is a chore now —random.random()spelled as CPython spells it andnext_doubledeleted with the regen; cells 2–5 are P61 ledger rows, and this issue stays open as therandomrow of the #2283 ledger.Internals (grep-only @ f8c37f9):
spy/random_module.spyimportsRandom as SysRandomand definesnext_double; the generated C# isRandomModuleModule.NextDouble();docs/stdlib/random.md:15,Spy/random/random_tests.spy,random_additional2_tests.spyandStdlibModuleGlobalConcurrencyTestsspellnext_double. Hypothesis for the chore: the .NET name was chosen because the module aliasesRandom; the emitter spellsglobal::System.Random, sodef random()should compile — the chore measures. Mechanism: rename at rung 1, deletenext_double(one way to do it, CLAUDE.md anti-patterns), regenerateRandom/Random.cs, the docs page and the spy tests in one commit;StdlibModuleGlobalRosterTestsunchanged (no new global). Cells 2–5 (theRandomclass andSystemRandom,getstate/setstate—System.Randomexposes no state, a deviation candidate —randbytes,triangular, the nine*variatefunctions) are ruled per name by the #2283 ledger as implemented at rung 1 or adocs/deviations.yamlrow; seeded sequences differing from CPython stay a documented deviation, not a cell. Fidelity scope: cell 1 is a Sharpy convention (expose the Python API), not a python-parity aperture; the name set of a stdlib module is an unlisted surface (R-CM) that P61 exists to rule per module. Rejected: (b) promoterandom.__all__and implement all 26 names now (pre-empts the ledger); (c) deviation rows for cells 2–5 now and close. The rename is one-way for C# consumers of the generated module (R-GW: no users before 1.0); (a) stacks under (b). Lane / reach: false-refusal · program. Class tracker: this issue.fix(random): the module function is random.random(), as CPython spell…
antonsynd commentedon Oct 22, 2026
Landed (chores band 2026-10-22) on
dev@ 4d22212, PR #2346 to mainline. Whole-solution gate 81669/0/7 @ 4d22212 (measured), against 81093/0/7 @ be4664e.Cell 1, commit 43f85a9. The function is
random.random(), andnext_doubleis deleted.Stays open for the rest of the class: the
Randomclass,getstate/setstate,randbytes,triangularand the*variatefunctions.New cell (agent probe): did-you-mean suggests the private
_random.Merge pull request #2346 from antonsynd/dev
antonsynd commentedon Oct 22, 2026
Batching 2026-10-22 @ 4d22212 — charter
.claude/plans/remediation-round-2026-09-20-batching.md— pointer only. The did-you-mean cell posted here by the chores-band agent (random.randm()→ "Did you mean '_random'?") is re-homed to #2096 (P25, the did-you-mean candidate-set class; cell comment there @ be4664e). Cells 2–5 stay P61 ledger rows under R-HL; this issue's placement is unchanged.