[Autoloop: build-tsb-pandas-typescript-migration] Experiment Log 2026-04

[github-actions]

🤖 Autoloop — an iterative optimization agent for this repository.

Program

Goal: Build tsb, a complete TypeScript port of pandas, one feature at a time.
Target files: src/**, tests/**, playground/**, and config files
Metric: pandas_features_ported (higher is better)
Current best: 14 (established in Iteration 13)

Iteration History

Iteration 13 — 2026-04-04 08:47 UTC — Run

• Status: ✅ Accepted
• Change: src/core/sort.ts — sorting utilities: nlargest/nsmallest (Series & DataFrame, keep='first'/'last'/'all'), rank (5 methods: average/min/max/first/dense, naOption, pct), rankDataFrame. 40+ unit + 5 property tests.
• Metric: 14 (previous best: 13, delta: +1)
• Commit: 688f640

Iteration 12 — 2026-04-04 08:25 UTC — Run

• Status: ✅ Accepted
• Change: src/core/datetime.ts — Series.dt DateTime accessor: calendar components, ISO week, dayofweek (Mon=0), strftime (17 directives). 78 unit + 6 property tests.
• Metric: 13 (previous best: 12, delta: +1)
• Commit: 6ce8c17

Iteration 11 — 2026-04-04 07:48 UTC — Run

• Status: ✅ Accepted
• Change: src/core/missing.ts — full missing-data utility suite: isna/notna/isnull/notnull, ffill/bfill (with limit), fillnaSeries/fillnaDataFrame, dropnaSeries, dropnaDataFrame, interpolate/interpolateDataFrame.
• Metric: 12 (previous best: 11, delta: +1)
• Commit: 4c2a1ea

Iteration 10 — 2026-04-04 07:30 UTC — Run

• Status: ✅ Accepted | Metric: 11 (+1) | Commit: 40058db
• Change: merge() — database-style joins inner/left/right/outer, on/left_on/right_on, many-to-many.

Iteration 9 — 2026-04-04 06:49 UTC — Run

• Status: ✅ Accepted | Metric: 10 (+1) | Commit: 6bd3f36
• Change: StringAccessor (src/core/strings.ts) — 20+ vectorized string methods via series.str.

Iteration 8 — 2026-04-04 06:24 UTC — Run

• Status: ✅ Accepted | Metric: 9 (+1) | Commit: 6fb9189
• Change: Index-aligned arithmetic — src/core/ops.ts. DataFrame.add/sub/mul/div/floordiv/mod/pow.

Iteration 7 — 2026-04-04 05:55 UTC — Run

• Status: ✅ Accepted | Metric: 8 (+1) | Commit: ee507e5
• Change: concat() — axis=0/1, outer/inner join, ignoreIndex

Iteration 6 — 2026-04-04 05:13 UTC — Run

• Status: ✅ Accepted | Metric: 7 (+1) | Commit: 57d00f3
• Change: GroupBy — DataFrameGroupBy + SeriesGroupBy with full split-apply-combine API

Iterations 1–5

• Iter 5 ✅ DataFrame (metric=6) · Iter 4 ⚠️ Error · Iter 3 ✅ Dtype+Series (metric=5) · Iter 2 ✅ Dtype+Index (metric=4) · Iter 1 ✅ Foundation (metric=1)

Generated by Autoloop · ● 8.4M · ◷

[github-actions]

---

Warning

The create_pull_request operation failed: Failed to apply patch. The code changes were not applied.

Iteration 2 — 2026-04-03 19:10 UTC — Run

• Status: ✅ Accepted
• Change: Implemented Dtype system — immutable singleton descriptors for all 16 pandas dtypes with kind, itemsize, type predicates, safe casting, commonType(), and Dtype.inferFrom(). Also fixed pre-existing noUncheckedIndexedAccess errors in base-index.ts and widened Index(T) signatures.
• Metric: 4 (previous best: 1, delta: +3)
• Commit: a45d5c1

Generated by Autoloop · ● 6.6M · ◷

[github-actions]

---

Note

The pull request was not created — a fallback review issue was created instead due to protected file changes: #9

🤖 Updated experiment log.

Iteration History

Iteration 3 — 2026-04-04 01:25 UTC — Run

• Status: ✅ Accepted
• Change: Implemented Dtype system (16 singletons with kind, itemsize, casting, inferFrom) and Series<T> (1-D labeled array with arithmetic, statistical, sorting, masking, and missing-value methods)
• Metric: 5 (previous best: 4, delta: +1)
• Commit: 36e76a5

Iteration 2 — 2026-04-03 19:10 UTC — Run

• Status: ✅ Accepted
• Change: Index type system (base-index.ts, range-index.ts) — previous Copilot agent work merged into autoloop branch; dtype.ts stub
• Metric: 4 (previous best: 1, delta: +3)

Iteration 1 — 2026-04-03 16:54 UTC — Run

• Status: ✅ Accepted
• Change: Complete project foundation (package.json, tsconfig, biome.json, CI/CD, playground, AGENTS.md, types.ts)
• Metric: 1 (baseline)

Current best: pandas_features_ported = 5

Generated by Autoloop · ● 8.7M · ◷

[github-actions]

---

Warning

The create_pull_request operation failed: Failed to apply patch. The code changes were not applied.

Iteration 4 — 2026-04-04 03:55 UTC — Run

• Status: ✅ Accepted
• Change: Implemented DataFrame — column-oriented 2-D labeled table with multiple constructors, loc/iloc indexers, head/tail, assign/drop/select/rename, isna/notna/dropna/fillna, filter, aggregations (sum/mean/min/max/std/count/describe), sortValues/sortIndex, apply, iteritems/iterrows, toRecords/toDict/toArray, resetIndex/setIndex, toString
• Metric: 6 (previous best: 5, delta: +1)
• Commit: 34e367a

Generated by Autoloop · 🤖

Generated by Autoloop · ● 12.2M · ◷

[github-actions]

Iteration 5 — 2026-04-04 04:58 UTC — Run

• Status: ✅ Accepted
• Change: Implemented DataFrame — 2-D column-oriented labeled table. fromColumns/fromRecords/from2D constructors; shape/ndim/size/empty; col/get/has; head/tail/iloc/loc; assign/drop/select/rename; isna/notna/dropna/fillna; filter; sum/mean/min/max/std/count/describe; sortValues/sortIndex; apply(axis=0/1); items/iterrows; toRecords/toDict/toArray; resetIndex/setIndex; toString.
• Metric: 6 (previous best: 5, delta: +1)
• Notes: Previous iteration 4 (run 23970468437) implemented DataFrame but PR creation failed — this run successfully commits the same feature. Biome useLiteralKeys vs TypeScript noPropertyAccessFromIndexSignature conflict resolved via toEqual in tests. noExcessiveCognitiveComplexity addressed by extracting compareScalarPair and computeColumnStats helpers.

Generated by Autoloop · ● 7.5M · ◷

[github-actions]

Iteration 6 — 2026-04-04 05:20 UTC — Run

• Status: ✅ Accepted
• Change: Implemented SeriesGroupBy and DataFrameGroupBy — the split-apply-combine engine with sum, mean, min, max, count, first, last, std, agg(func|name|spec), transform, apply, getGroup, Symbol.iterator. 35+ tests + 4 property-based tests.
• Metric: 7 (previous best: 6, delta: +1)
• Commit: 8aa0900
• Notes: GroupBy builds on top of the DataFrame implementation (iteration 5). Key design decision: group keys are Label (not Scalar) to allow clean Index construction. Multi-column keys are tab-separated composite strings. toLabel() helper safely coerces bigint/Date to strings.

Current best metric: pandas_features_ported = 7 (established in iteration 6)

Generated by Autoloop · ● 8.7M · ◷

[github-actions]

Iteration 6 (run 23972003902) — 2026-04-04 05:13 UTC — Run

• Status: ✅ Accepted
• Change: Implemented GroupBy — DataFrameGroupBy + SeriesGroupBy with full split-apply-combine engine. All aggregations: sum, mean, min, max, count, std, first, last, size. agg() with named/fn/per-column specs. transform(), apply(), filter(), getGroup(). Multi-key groupby. DataFrame.groupby() + Series.groupby().
• Metric: 7 (previous best: 6, delta: +1)
• Commit: 57d00f3

Generated by Autoloop · ● 9.9M · ◷

[github-actions]

Iteration 8 — 2026-04-04 06:24 UTC — Run

• Status: ✅ Accepted
• Change: Implemented index-aligned arithmetic — src/core/ops.ts (alignSeries, alignedBinaryOp, alignDataFrames, alignedDataFrameBinaryOp). Updated Series._scalarOp for index alignment. Added DataFrame.add/sub/mul/div/floordiv/mod/pow.
• Metric: 9 (previous best: 8, delta: +1)
• Commit: 6fb9189

Generated by Autoloop · ● 10.6M · ◷

[github-actions]

Iteration 10 — 2026-04-04 07:30 UTC — Run

• Status: ✅ Accepted
• Change: merge() — database-style joins (inner/left/right/outer), on/left_on/right_on, suffixes, many-to-many
• Metric: 11 (previous best: 10, delta: +1)
• Commit: 40058db
• PR: #aw_pr10

Generated by Autoloop · ● 5.1M · ◷

[github-actions]

Iteration 12 — 2026-04-04 08:25 UTC — Run

• Status: ✅ Accepted
• Change: src/core/datetime.ts — Series.dt DateTime accessor: year/month/day/hour/minute/second/millisecond, quarter, dayofyear, week, dayofweek (Mon=0), boundary flags (is_month_start/end, is_quarter_start/end, is_year_start/end, is_leap_year, days_in_month), strftime, date(), time(), normalize(), total_seconds(), timestamp_ms(). 78 unit tests + 6 property-based tests.
• Metric: 13 (previous best: 12, delta: +1)
• Commit: 6ce8c17

Generated by Autoloop · ● 6.1M · ◷

[github-actions]

Iteration 15 — 2026-04-04 09:52 UTC — Run

• Status: ✅ Accepted
• Change: src/core/compare.ts — element-wise eq/ne/lt/gt/le/ge for Series & DataFrame, logical and/or/xor/not, anySeries/allSeries/anyDataFrame/allDataFrame. 45 unit tests + 4 property-based tests, 100% coverage.
• Metric: 16 (previous best: 15, delta: +1)
• Commit: d49f29e

Generated by Autoloop · ● 6.4M · ◷

[github-actions]

Iteration 16 — 2026-04-04 10:12 UTC — Run

• Status: ✅ Accepted
• Change: src/reshape/ — pivot/pivotTable (long→wide with 7 aggfuncs), melt (wide→long unpivot), stack/unstack (column↔row collapse without MultiIndex)
• Metric: 19 (previous best: 16, delta: +3)
• Commit: 075baf8

Generated by Autoloop · ● 9.8M · ◷

[github-actions]

Iteration 17 — 2026-04-04 10:53 UTC — Run

• Status: ✅ Accepted
• Change: src/window/ — rolling (SeriesRolling/DataFrameRolling), expanding (SeriesExpanding/DataFrameExpanding), ewm (SeriesEWM/DataFrameEWM). 74 unit + 4 property-based tests.
• Metric: 22 (previous best: 19, delta: +3)
• Commit: c0b018f

Generated by Autoloop · ● 7.5M · ◷