fix lake count (54,667) + breakdown

docs/schema.md

@@ -1,143 +1,80 @@
 # `logo-lab/dlte` per-config schema reference
 
-## `lake` config (5 splits)
+Two configs:
 
-### `parents_tabfact`, `parents_wtq`
+- **`manifests`**: lightweight metadata as one parquet stream with a unified
+  `(record_type, record_json)` schema.
+- **`lake`**: full table content (parents + fragments + distractors) with
+  typed columns + provenance keys.
 
-| Column | Type | Notes |
-|---|---|---|
-| `table_id` | string | original parent identifier (TabFact: `r-N`; WTQ: csv id) |
-| `source` | string | `tabfact` or `wtq` |
-| `n_rows` | int32 | parent row count |
-| `n_cols` | int32 | parent column count |
-| `header` | list<string> | column names |
-| `table_csv` | large_string | full parent serialized as CSV |
-| `provenance_url` | string | upstream link |
-| `license_tag` | string | per-row license inheritance |
-
-### `targets`
+## `manifests` config
 
-| Column | Type | Notes |
-|---|---|---|
-| `target_id` | string | unique fragment id |
-| `parent_id` | string | links to parents_*.table_id |
-| `parent_source` | string | `tabfact` or `wtq` |
-| `kind` | string | `union` or `join` |
-| `noise_tier` | string | `clean` / `schema` / `cell` / `hard` |
-| `removed_rows` | list<int32> | indices of rows removed from parent (only for `union`) |
-| `removed_cols` | list<int32> | indices of cols removed from parent (only for `join`) |
-| `table_csv` | large_string | the fragment table |
-| `license_tag` | string | inherited |
-
-### `distractors`
+Single split (`train`), 109,336 records.
 
 | Column | Type | Notes |
 |---|---|---|
-| `distractor_id` | string | matches `manifests/ckan_distractor_ids` |
-| `source` | string | always `lakebench_ckan` |
-| `table_csv` | large_string | the CKAN distractor table |
-| `license_tag` | string | `cc-by-4.0 (LakeBench Zenodo); per-portal upstream` |
+| `record_type` | `string` | one of: `parent`, `fragment`, `lake_table`, `split_assignment`, `query_task`, `ckan_distractor`, `fragmentation_config`, `gt_validation` |
+| `record_json` | `string` | JSON-encoded record content (variable schema per `record_type`) |
 
-### `queries`
+### `record_type` counts
 
-| Column | Type | Notes |
+| record_type | rows | source |
 |---|---|---|
-| `query_id` | string | seed query identifier |
-| `parent_id` | string | parent the query was derived from |
-| `parent_source` | string | `tabfact` or `wtq` |
-| `noise_tier` | string | |
-| `seed_csv` | large_string | the seed query (parent with rows + cols removed) |
-| `removed_rows` | list<int32> | the union target's rows |
-| `removed_cols` | list<int32> | the join target's cols |
-| `license_tag` | string | inherited |
-
-## `manifests` config (8 splits)
-
-### `splits`
+| `lake_table` | 47,772 | `manifests/lake_manifest.jsonl` |
+| `ckan_distractor` | 36,740 | `manifests/ckan_distractor_ids.txt` (one row per id) |
+| `fragment` | 16,548 | `manifests/fragments_manifest.jsonl` |
+| `query_task` | 5,516 | `ground_truth/query_tasks.jsonl` |
+| `parent` | 1,379 | `manifests/parents_filtered.jsonl` |
+| `split_assignment` | 1,379 | `manifests/splits.json` (one row per (parent_id, split)) |
+| `fragmentation_config` | 1 | `config/fragmentation.json` |
+| `gt_validation` | 1 | `ground_truth/gt_validation.json` |
+
+### Record schemas (after `json.loads(record_json)`)
+
+- **`parent`**: `{parent_id, dataset, csv_path, csv_stem, table_id_in_embeddings, n_rows, n_cols, key_col, key_col_idx, key_uniqueness, key_non_null_ratio, key_is_text}`
+- **`fragment`**: `{table_id, parent_id, split, role, fragment_type, noise_tier, replicate_id, csv_path, n_rows, n_cols, key_col_fragment, key_col_parent}`
+- **`lake_table`**: `{table_id, source, parent_id, fragment_type, noise_tier, replicate_id, split, csv_path, n_rows, n_cols}` — the full lake of 47,772 (= 11,032 targets + 36,740 distractors)
+- **`split_assignment`**: `{parent_id, split}` (split ∈ {train, dev, test})
+- **`query_task`**: `{query_table_id, parent_id, split, noise_tier, replicate_id, relevant: [{table_id, relation}, ...]}`
+- **`ckan_distractor`**: `{distractor_id}`
+- **`fragmentation_config`**: full config blob (tiers, noise specs)
+- **`gt_validation`**: validation ground truth blob
+
+## `lake` config
+
+Single split (`train`), **54,667 rows** (= 1,379 parents + 16,548 fragments [5,516 query_seed + 5,516 target_union + 5,516 target_join] + 36,740 distractors). Sharded by byte budget (~1.5 GB raw per shard) so each parquet column page stays under the INT32 limit.
 
 | Column | Type | Notes |
 |---|---|---|
-| `parent_id` | string | |
-| `split` | string | `train` / `dev` / `test` |
-| `parent_source` | string | |
-
-(827 train / 207 dev / 345 test rows)
-
-### `parents_filtered`
-
-| Column | Type | Notes |
+| `table_id` | `string` | unique within the lake |
+| `kind` | `string` | `parent_tabfact` / `parent_wtq` / `target_union` / `target_join` / `query_seed` / `distractor` |
+| `parent_id` | `string` | empty for distractors |
+| `parent_source` | `string` | `tabfact` / `wtq` / empty (distractor) |
+| `noise_tier` | `int32` | 0/1/2/3 for fragments; -1 for parents/distractors |
+| `fragment_type` | `string` | `seed` / `union` / `join` / empty |
+| `split` | `string` | `train` / `dev` / `test` / empty (distractors are unsplit) |
+| `csv_text` | `large_string` | raw CSV content of the table |
+| `n_rows` | `int32` | row count (excluding header) |
+| `n_cols` | `int32` | column count |
+
+### Counts by `kind`
+
+| kind | count | provenance |
 |---|---|---|
-| `parent_id` | string | |
-| `parent_source` | string | |
-| `n_rows` | int32 | |
-| `n_cols` | int32 | |
-| `passes_filter` | bool | should always be `true` since this is the post-filter list |
-| `provenance_url` | string | |
+| `parent_tabfact` | 989 | TabFact (Chen et al., ICLR 2020) |
+| `parent_wtq` | 390 | WikiTableQuestions (Pasupat & Liang, ACL 2015) |
+| `target_union` | 5,516 | derived: row-blocks removed from parents |
+| `target_join` | 5,516 | derived: column-blocks removed from parents |
+| `query_seed` | 5,516 | derived: parent minus removed rows + columns |
+| `distractor` | 36,740 | LakeBench-Srinivas Zenodo CKAN Subset pool |
+| **Total** | **54,667** | — |
 
-(1,379 rows)
+Of these, **47,772 = `target_union` + `target_join` + `distractor`** is the *retrieval lake* the pipeline searches over. The other 6,895 (1,379 parents + 5,516 query seeds) are released alongside as pipeline I/O.
 
-### `fragments` (16,548 rows)
+## Cross-config keys
 
-| Column | Type | Notes |
-|---|---|---|
-| `fragment_id` | string | |
-| `parent_id` | string | |
-| `parent_source` | string | |
-| `kind` | string | `seed` / `union_target` / `join_target` |
-| `noise_tier` | string | |
-| `provenance` | string | JSON-encoded extra context |
+- `manifests.lake_table.record_json["table_id"]` ↔ `lake.table_id` for `kind ∈ {target_union, target_join, distractor}`
+- `manifests.parent.record_json["parent_id"]` ↔ `lake.parent_id` (and matches `lake.table_id` for `kind ∈ {parent_*}`)
+- `manifests.split_assignment.record_json["parent_id"]` ↔ `lake.parent_id` for parent rows
 
-### `lake` (the retrieval lake of exactly 47,772 tables)
-
-This split lists only the tables that are searchable in Stage-1 retrieval —
-**i.e., union/join targets + CKAN distractors**. Parents and query seeds are
-pipeline I/O (kept in the `lake` config's other splits and the `manifests`
-config, but not in *this* split).
-
-| Column | Type | Notes |
-|---|---|---|
-| `table_id` | string | |
-| `kind` | string | `union_target` / `join_target` / `distractor` (exactly these three) |
-| `parent_id` | string | nullable; only set for derivatives |
-| `parent_source` | string | nullable; `tabfact` or `wtq` for derivative targets |
-| `noise_tier` | string | nullable; `clean` / `schema` / `cell` / `hard` for targets only |
-
-### `ckan_distractor_ids`
-
-| Column | Type | Notes |
-|---|---|---|
-| `distractor_id` | string | (36,740 rows; matches `lake.distractors.distractor_id`) |
-
-### `query_tasks`
-
-| Column | Type | Notes |
-|---|---|---|
-| `query_id` | string | |
-| `parent_id` | string | |
-| `parent_source` | string | |
-| `noise_tier` | string | |
-| `task_type` | string | one of `union_only` / `join_only` / `union_and_join` |
-| `gt_union_target_ids` | list<string> | ground-truth union target fragment ids |
-| `gt_join_target_ids` | list<string> | ground-truth join target fragment ids |
-
-### `gt_validation`
-
-Per-query validation rows used for the dev-split macro-F1 calibration of
-Stage-2 thresholds.
-
-| Column | Type | Notes |
-|---|---|---|
-| `query_id` | string | |
-| `gt_action` | string | `union` / `join` / `none` |
-| `gt_target_id` | string | nullable |
-| `metadata` | string | JSON-encoded |
-
-### `table_maps`
-
-| Column | Type | Notes |
-|---|---|---|
-| `table_id` | string | |
-| `kind` | string | |
-| `parent_id` | string | |
-| `csv_path` | string | path within `lake/*` parquet (informational only — actual csv lives in `lake` config) |
-| `stage2_calibrated_thresholds` | string | JSON-encoded; only populated for the (Stage-1 backbone, Stage-2 model) calibration result rows used in the paper |
+Stage-2 calibrated thresholds for the (Stage-1 backbone, Stage-2 column model) pairs evaluated in the paper are not included in this release; they are deterministic given the dev split and reproducible from the operator spec in the paper appendix.