@secondlayer/shared 13.13.0 → 13.14.0

package/dist/src/archive/copy-loader.d.ts

@@ -1,20 +1,19 @@
 import { Writable } from "node:stream";
 /**
-* Streaming `COPY ... FROM STDIN` loader for canonical archive rows.
+* The canonical archive's datasets — the one list every exporter, loader,
+* verifier and fetch gate derives from.
 *
-* Restoring history row-by-row is not viable at chain scale: batched INSERTs
-* measured ~2k rows/s on this schema, which is days for a full genesis restore.
-* COPY moves the same data at ~64k rows/s because it is one statement per
-* partition with no per-row round-trip.
+* Order is load order: each dataset's foreign keys point only at datasets
+* earlier in the list, so loading in this order never violates a constraint.
 *
-* This module deliberately knows nothing about Parquet or about where rows come
-* from — it accepts the archive's row shape as an async iterable. That keeps the
-* heavy file-format dependency out of every consumer that only needs to write
-* rows, and lets the CLI and the indexer share one loader rather than growing
-* two subtly different CSV encoders.
+* Per-dataset code branches with an exhaustive `switch` ending in
+* `assertNeverDataset`, never a trailing default: adding a dataset here must
+* fail to compile everywhere it needs its own schema, SQL or table, instead of
+* silently inheriting another dataset's.
 */
-/** The three archive datasets: blocks, transactions, and events. */
-type ArchiveDataset = "blocks" | "transactions" | "events";
+declare const ARCHIVE_DATASETS: readonly ["blocks", "transactions", "events", "state_writes"];
+/** One archive dataset: blocks, transactions, events or state writes. */
+type ArchiveDataset = (typeof ARCHIVE_DATASETS)[number];
 /** Row shapes exactly as the archive's Parquet columns are named. */
 type ArchiveRow = Record<string, unknown>;
 /**

package/dist/src/archive/copy-loader.js

@@ -1,5 +1,47 @@
 // src/archive/copy-loader.ts
 import { once } from "node:events";
+
+// src/archive/datasets.ts
+var ARCHIVE_DATASETS = [
+  "blocks",
+  "transactions",
+  "events",
+  "state_writes"
+];
+function isArchiveDataset(value) {
+  return ARCHIVE_DATASETS.includes(value);
+}
+var OPTIONAL_DATASET = {
+  blocks: false,
+  transactions: false,
+  events: false,
+  state_writes: true
+};
+function isOptionalDataset(dataset) {
+  return OPTIONAL_DATASET[dataset];
+}
+function manifestPublishes(manifest, dataset) {
+  if (!isOptionalDataset(dataset))
+    return true;
+  return (manifest.partitions ?? []).some((p) => p.dataset === dataset) || (manifest.zero_record_ranges ?? []).some((z) => z.dataset === dataset);
+}
+function archiveHeightColumn(dataset) {
+  switch (dataset) {
+    case "blocks":
+      return "height";
+    case "transactions":
+    case "events":
+    case "state_writes":
+      return "block_height";
+    default:
+      return assertNeverDataset(dataset);
+  }
+}
+function assertNeverDataset(dataset) {
+  throw new Error(`unhandled archive dataset: ${String(dataset)}`);
+}
+
+// src/archive/copy-loader.ts
 function csvField(value) {
   if (value === null || value === undefined)
     return "\\N";
@@ -9,6 +51,9 @@ function csvField(value) {
 function asNumber(value) {
   return typeof value === "bigint" ? Number(value) : Number(value);
 }
+function asOptionalNumber(value) {
+  return value === null || value === undefined ? null : asNumber(value);
+}
 function asText(value) {
   return Buffer.isBuffer(value) ? value.toString("utf8") : String(value);
 }
@@ -16,48 +61,64 @@ function asOptionalText(value) {
   return value === null || value === undefined ? null : asText(value);
 }
 function toCsvLine(dataset, row) {
-  if (dataset === "blocks") {
-    return [
-      csvField(asNumber(row.height)),
-      csvField(asText(row.hash)),
-      csvField(asText(row.parent_hash)),
-      csvField(asNumber(row.burn_block_height)),
-      csvField(asOptionalText(row.burn_block_hash)),
-      csvField(asOptionalText(row.index_block_hash)),
-      csvField(asNumber(row.timestamp))
-    ].join(",");
-  }
-  if (dataset === "transactions") {
-    return [
-      csvField(asText(row.tx_id)),
-      csvField(asNumber(row.block_height)),
-      csvField(asNumber(row.tx_index)),
-      csvField(asText(row.type)),
-      csvField(asText(row.sender)),
-      csvField(asText(row.status)),
-      csvField(asOptionalText(row.contract_id)),
-      csvField(asOptionalText(row.function_name)),
-      csvField(asOptionalText(row.function_args_json)),
-      csvField(asOptionalText(row.raw_result)),
-      csvField(asText(row.raw_tx))
-    ].join(",");
+  switch (dataset) {
+    case "blocks":
+      return [
+        csvField(asNumber(row.height)),
+        csvField(asText(row.hash)),
+        csvField(asText(row.parent_hash)),
+        csvField(asNumber(row.burn_block_height)),
+        csvField(asOptionalText(row.burn_block_hash)),
+        csvField(asOptionalText(row.index_block_hash)),
+        csvField(asNumber(row.timestamp))
+      ].join(",");
+    case "transactions":
+      return [
+        csvField(asText(row.tx_id)),
+        csvField(asNumber(row.block_height)),
+        csvField(asNumber(row.tx_index)),
+        csvField(asText(row.type)),
+        csvField(asText(row.sender)),
+        csvField(asText(row.status)),
+        csvField(asOptionalText(row.contract_id)),
+        csvField(asOptionalText(row.function_name)),
+        csvField(asOptionalText(row.function_args_json)),
+        csvField(asOptionalText(row.raw_result)),
+        csvField(asText(row.raw_tx))
+      ].join(",");
+    case "events":
+      return [
+        csvField(asText(row.tx_id)),
+        csvField(asNumber(row.block_height)),
+        csvField(asNumber(row.event_index)),
+        csvField(asText(row.event_type)),
+        csvField(asText(row.data_json))
+      ].join(",");
+    case "state_writes":
+      return [
+        csvField(asNumber(row.block_height)),
+        csvField(asNumber(row.ordinal)),
+        csvField(asOptionalNumber(row.tx_index)),
+        csvField(asText(row.key)),
+        csvField(asText(row.value_hex))
+      ].join(",");
+    default:
+      return assertNeverDataset(dataset);
   }
-  return [
-    csvField(asText(row.tx_id)),
-    csvField(asNumber(row.block_height)),
-    csvField(asNumber(row.event_index)),
-    csvField(asText(row.event_type)),
-    csvField(asText(row.data_json))
-  ].join(",");
 }
 function copyStatement(dataset) {
-  if (dataset === "blocks") {
-    return "COPY blocks (height, hash, parent_hash, burn_block_height, burn_block_hash, index_block_hash, timestamp) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
-  }
-  if (dataset === "transactions") {
-    return "COPY transactions (tx_id, block_height, tx_index, type, sender, status, contract_id, function_name, function_args, raw_result, raw_tx) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
+  switch (dataset) {
+    case "blocks":
+      return "COPY blocks (height, hash, parent_hash, burn_block_height, burn_block_hash, index_block_hash, timestamp) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
+    case "transactions":
+      return "COPY transactions (tx_id, block_height, tx_index, type, sender, status, contract_id, function_name, function_args, raw_result, raw_tx) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
+    case "events":
+      return "COPY events (tx_id, block_height, event_index, type, data) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
+    case "state_writes":
+      return "COPY state_writes (block_height, ordinal, tx_index, key, value_hex) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
+    default:
+      return assertNeverDataset(dataset);
   }
-  return "COPY events (tx_id, block_height, event_index, type, data) FROM STDIN WITH (FORMAT csv, NULL '\\N')";
 }
 async function writeRowsToCopyStream(params) {
   const { writable, dataset, rows } = params;
@@ -91,5 +152,5 @@ export {
   writeRowsToCopyStream
 };
 
-//# debugId=4643C93B767098ED64756E2164756E21
+//# debugId=63CC074406EEB2EC64756E2164756E21
 //# sourceMappingURL=copy-loader.js.map

package/dist/src/archive/copy-loader.js.map

@@ -1,10 +1,11 @@
 {
   "version": 3,
-  "sources": ["../src/archive/copy-loader.ts"],
+  "sources": ["../src/archive/copy-loader.ts", "../src/archive/datasets.ts"],
   "sourcesContent": [
-    "import { once } from \"node:events\";\nimport type { Writable } from \"node:stream\";\n\n/**\n * Streaming `COPY ... FROM STDIN` loader for canonical archive rows.\n *\n * Restoring history row-by-row is not viable at chain scale: batched INSERTs\n * measured ~2k rows/s on this schema, which is days for a full genesis restore.\n * COPY moves the same data at ~64k rows/s because it is one statement per\n * partition with no per-row round-trip.\n *\n * This module deliberately knows nothing about Parquet or about where rows come\n * from — it accepts the archive's row shape as an async iterable. That keeps the\n * heavy file-format dependency out of every consumer that only needs to write\n * rows, and lets the CLI and the indexer share one loader rather than growing\n * two subtly different CSV encoders.\n */\n\n/** The three archive datasets: blocks, transactions, and events. */\nexport type ArchiveDataset = \"blocks\" | \"transactions\" | \"events\";\n\n/** Row shapes exactly as the archive's Parquet columns are named. */\nexport type ArchiveRow = Record<string, unknown>;\n\n/**\n * CSV encoding for `WITH (FORMAT csv, NULL '\\N')`: every present value is\n * quoted with embedded quotes doubled, so a value that happens to equal the\n * NULL marker cannot be misread as SQL NULL. Only a genuinely absent value is\n * written unquoted as `\\N`.\n */\nexport function csvField(value: string | number | null | undefined): string {\n\tif (value === null || value === undefined) return \"\\\\N\";\n\tconst text = typeof value === \"number\" ? String(value) : value;\n\treturn `\"${text.replace(/\"/g, '\"\"')}\"`;\n}\n\nfunction asNumber(value: unknown): number {\n\treturn typeof value === \"bigint\" ? Number(value) : Number(value as number);\n}\n\nfunction asText(value: unknown): string {\n\treturn Buffer.isBuffer(value) ? value.toString(\"utf8\") : String(value);\n}\n\nfunction asOptionalText(value: unknown): string | null {\n\treturn value === null || value === undefined ? null : asText(value);\n}\n\nexport function toCsvLine(dataset: ArchiveDataset, row: ArchiveRow): string {\n\tif (dataset === \"blocks\") {\n\t\treturn [\n\t\t\tcsvField(asNumber(row.height)),\n\t\t\tcsvField(asText(row.hash)),\n\t\t\tcsvField(asText(row.parent_hash)),\n\t\t\tcsvField(asNumber(row.burn_block_height)),\n\t\t\tcsvField(asOptionalText(row.burn_block_hash)),\n\t\t\tcsvField(asOptionalText(row.index_block_hash)),\n\t\t\tcsvField(asNumber(row.timestamp)),\n\t\t].join(\",\");\n\t}\n\tif (dataset === \"transactions\") {\n\t\treturn [\n\t\t\tcsvField(asText(row.tx_id)),\n\t\t\tcsvField(asNumber(row.block_height)),\n\t\t\tcsvField(asNumber(row.tx_index)),\n\t\t\tcsvField(asText(row.type)),\n\t\t\tcsvField(asText(row.sender)),\n\t\t\tcsvField(asText(row.status)),\n\t\t\tcsvField(asOptionalText(row.contract_id)),\n\t\t\tcsvField(asOptionalText(row.function_name)),\n\t\t\tcsvField(asOptionalText(row.function_args_json)),\n\t\t\tcsvField(asOptionalText(row.raw_result)),\n\t\t\tcsvField(asText(row.raw_tx)),\n\t\t].join(\",\");\n\t}\n\treturn [\n\t\tcsvField(asText(row.tx_id)),\n\t\tcsvField(asNumber(row.block_height)),\n\t\tcsvField(asNumber(row.event_index)),\n\t\tcsvField(asText(row.event_type)),\n\t\tcsvField(asText(row.data_json)),\n\t].join(\",\");\n}\n\n/**\n * `canonical` is omitted from the blocks column list: every archived block is\n * canonical by definition, and the schema defaults it to true, so carrying it\n * through every row would be dead weight.\n */\nexport function copyStatement(dataset: ArchiveDataset): string {\n\tif (dataset === \"blocks\") {\n\t\treturn \"COPY blocks (height, hash, parent_hash, burn_block_height, burn_block_hash, index_block_hash, timestamp) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n\t}\n\tif (dataset === \"transactions\") {\n\t\treturn \"COPY transactions (tx_id, block_height, tx_index, type, sender, status, contract_id, function_name, function_args, raw_result, raw_tx) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n\t}\n\treturn \"COPY events (tx_id, block_height, event_index, type, data) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n}\n\n/**\n * Stream rows onto an open COPY writable. The caller owns opening it (drivers\n * differ in how a COPY stream is obtained) and this owns the wire format,\n * backpressure, and failure semantics.\n *\n * On error the stream is destroyed rather than ended, so Postgres aborts the\n * COPY instead of committing a partial one — an interrupted load must leave\n * nothing behind, not half a partition.\n */\nexport async function writeRowsToCopyStream(params: {\n\twritable: Writable;\n\tdataset: ArchiveDataset;\n\trows: AsyncIterable<ArchiveRow>;\n\tonProgress?: (rowsWritten: number) => void;\n\tprogressEvery?: number;\n}): Promise<number> {\n\tconst { writable, dataset, rows } = params;\n\tconst progressEvery = params.progressEvery ?? 500_000;\n\tlet count = 0;\n\ttry {\n\t\tfor await (const row of rows) {\n\t\t\tconst line = `${toCsvLine(dataset, row)}\\n`;\n\t\t\tif (!writable.write(line)) {\n\t\t\t\tawait once(writable, \"drain\");\n\t\t\t}\n\t\t\tcount++;\n\t\t\tif (params.onProgress && count % progressEvery === 0) {\n\t\t\t\tparams.onProgress(count);\n\t\t\t}\n\t\t}\n\t\tawait new Promise<void>((resolve, reject) => {\n\t\t\twritable.end((error?: Error | null) =>\n\t\t\t\terror ? reject(error) : resolve(),\n\t\t\t);\n\t\t});\n\t} catch (error) {\n\t\twritable.destroy(error instanceof Error ? error : new Error(String(error)));\n\t\tthrow error;\n\t}\n\treturn count;\n}\n"
+    "import { once } from \"node:events\";\nimport type { Writable } from \"node:stream\";\nimport { type ArchiveDataset, assertNeverDataset } from \"./datasets.ts\";\n\n/**\n * Streaming `COPY ... FROM STDIN` loader for canonical archive rows.\n *\n * Restoring history row-by-row is not viable at chain scale: batched INSERTs\n * measured ~2k rows/s on this schema, which is days for a full genesis restore.\n * COPY moves the same data at ~64k rows/s because it is one statement per\n * partition with no per-row round-trip.\n *\n * This module deliberately knows nothing about Parquet or about where rows come\n * from — it accepts the archive's row shape as an async iterable. That keeps the\n * heavy file-format dependency out of every consumer that only needs to write\n * rows, and lets the CLI and the indexer share one loader rather than growing\n * two subtly different CSV encoders.\n */\n\nexport type { ArchiveDataset } from \"./datasets.ts\";\n\n/** Row shapes exactly as the archive's Parquet columns are named. */\nexport type ArchiveRow = Record<string, unknown>;\n\n/**\n * CSV encoding for `WITH (FORMAT csv, NULL '\\N')`: every present value is\n * quoted with embedded quotes doubled, so a value that happens to equal the\n * NULL marker cannot be misread as SQL NULL. Only a genuinely absent value is\n * written unquoted as `\\N`.\n */\nexport function csvField(value: string | number | null | undefined): string {\n\tif (value === null || value === undefined) return \"\\\\N\";\n\tconst text = typeof value === \"number\" ? String(value) : value;\n\treturn `\"${text.replace(/\"/g, '\"\"')}\"`;\n}\n\nfunction asNumber(value: unknown): number {\n\treturn typeof value === \"bigint\" ? Number(value) : Number(value as number);\n}\n\nfunction asOptionalNumber(value: unknown): number | null {\n\treturn value === null || value === undefined ? null : asNumber(value);\n}\n\nfunction asText(value: unknown): string {\n\treturn Buffer.isBuffer(value) ? value.toString(\"utf8\") : String(value);\n}\n\nfunction asOptionalText(value: unknown): string | null {\n\treturn value === null || value === undefined ? null : asText(value);\n}\n\nexport function toCsvLine(dataset: ArchiveDataset, row: ArchiveRow): string {\n\tswitch (dataset) {\n\t\tcase \"blocks\":\n\t\t\treturn [\n\t\t\t\tcsvField(asNumber(row.height)),\n\t\t\t\tcsvField(asText(row.hash)),\n\t\t\t\tcsvField(asText(row.parent_hash)),\n\t\t\t\tcsvField(asNumber(row.burn_block_height)),\n\t\t\t\tcsvField(asOptionalText(row.burn_block_hash)),\n\t\t\t\tcsvField(asOptionalText(row.index_block_hash)),\n\t\t\t\tcsvField(asNumber(row.timestamp)),\n\t\t\t].join(\",\");\n\t\tcase \"transactions\":\n\t\t\treturn [\n\t\t\t\tcsvField(asText(row.tx_id)),\n\t\t\t\tcsvField(asNumber(row.block_height)),\n\t\t\t\tcsvField(asNumber(row.tx_index)),\n\t\t\t\tcsvField(asText(row.type)),\n\t\t\t\tcsvField(asText(row.sender)),\n\t\t\t\tcsvField(asText(row.status)),\n\t\t\t\tcsvField(asOptionalText(row.contract_id)),\n\t\t\t\tcsvField(asOptionalText(row.function_name)),\n\t\t\t\tcsvField(asOptionalText(row.function_args_json)),\n\t\t\t\tcsvField(asOptionalText(row.raw_result)),\n\t\t\t\tcsvField(asText(row.raw_tx)),\n\t\t\t].join(\",\");\n\t\tcase \"events\":\n\t\t\treturn [\n\t\t\t\tcsvField(asText(row.tx_id)),\n\t\t\t\tcsvField(asNumber(row.block_height)),\n\t\t\t\tcsvField(asNumber(row.event_index)),\n\t\t\t\tcsvField(asText(row.event_type)),\n\t\t\t\tcsvField(asText(row.data_json)),\n\t\t\t].join(\",\");\n\t\tcase \"state_writes\":\n\t\t\treturn [\n\t\t\t\tcsvField(asNumber(row.block_height)),\n\t\t\t\tcsvField(asNumber(row.ordinal)),\n\t\t\t\tcsvField(asOptionalNumber(row.tx_index)),\n\t\t\t\tcsvField(asText(row.key)),\n\t\t\t\tcsvField(asText(row.value_hex)),\n\t\t\t].join(\",\");\n\t\tdefault:\n\t\t\treturn assertNeverDataset(dataset);\n\t}\n}\n\n/**\n * `canonical` is omitted from the blocks column list: every archived block is\n * canonical by definition, and the schema defaults it to true, so carrying it\n * through every row would be dead weight.\n */\nexport function copyStatement(dataset: ArchiveDataset): string {\n\tswitch (dataset) {\n\t\tcase \"blocks\":\n\t\t\treturn \"COPY blocks (height, hash, parent_hash, burn_block_height, burn_block_hash, index_block_hash, timestamp) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n\t\tcase \"transactions\":\n\t\t\treturn \"COPY transactions (tx_id, block_height, tx_index, type, sender, status, contract_id, function_name, function_args, raw_result, raw_tx) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n\t\tcase \"events\":\n\t\t\treturn \"COPY events (tx_id, block_height, event_index, type, data) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n\t\tcase \"state_writes\":\n\t\t\treturn \"COPY state_writes (block_height, ordinal, tx_index, key, value_hex) FROM STDIN WITH (FORMAT csv, NULL '\\\\N')\";\n\t\tdefault:\n\t\t\treturn assertNeverDataset(dataset);\n\t}\n}\n\n/**\n * Stream rows onto an open COPY writable. The caller owns opening it (drivers\n * differ in how a COPY stream is obtained) and this owns the wire format,\n * backpressure, and failure semantics.\n *\n * On error the stream is destroyed rather than ended, so Postgres aborts the\n * COPY instead of committing a partial one — an interrupted load must leave\n * nothing behind, not half a partition.\n */\nexport async function writeRowsToCopyStream(params: {\n\twritable: Writable;\n\tdataset: ArchiveDataset;\n\trows: AsyncIterable<ArchiveRow>;\n\tonProgress?: (rowsWritten: number) => void;\n\tprogressEvery?: number;\n}): Promise<number> {\n\tconst { writable, dataset, rows } = params;\n\tconst progressEvery = params.progressEvery ?? 500_000;\n\tlet count = 0;\n\ttry {\n\t\tfor await (const row of rows) {\n\t\t\tconst line = `${toCsvLine(dataset, row)}\\n`;\n\t\t\tif (!writable.write(line)) {\n\t\t\t\tawait once(writable, \"drain\");\n\t\t\t}\n\t\t\tcount++;\n\t\t\tif (params.onProgress && count % progressEvery === 0) {\n\t\t\t\tparams.onProgress(count);\n\t\t\t}\n\t\t}\n\t\tawait new Promise<void>((resolve, reject) => {\n\t\t\twritable.end((error?: Error | null) =>\n\t\t\t\terror ? reject(error) : resolve(),\n\t\t\t);\n\t\t});\n\t} catch (error) {\n\t\twritable.destroy(error instanceof Error ? error : new Error(String(error)));\n\t\tthrow error;\n\t}\n\treturn count;\n}\n",
+    "/**\n * The canonical archive's datasets — the one list every exporter, loader,\n * verifier and fetch gate derives from.\n *\n * Order is load order: each dataset's foreign keys point only at datasets\n * earlier in the list, so loading in this order never violates a constraint.\n *\n * Per-dataset code branches with an exhaustive `switch` ending in\n * `assertNeverDataset`, never a trailing default: adding a dataset here must\n * fail to compile everywhere it needs its own schema, SQL or table, instead of\n * silently inheriting another dataset's.\n */\nexport const ARCHIVE_DATASETS = [\n\t\"blocks\",\n\t\"transactions\",\n\t\"events\",\n\t\"state_writes\",\n] as const;\n\n/** One archive dataset: blocks, transactions, events or state writes. */\nexport type ArchiveDataset = (typeof ARCHIVE_DATASETS)[number];\n\nexport function isArchiveDataset(value: string): value is ArchiveDataset {\n\treturn (ARCHIVE_DATASETS as readonly string[]).includes(value);\n}\n\n/**\n * Datasets a manifest may leave out. The exporter publishes `state_writes`\n * only from an instance that holds it from the start of the exported range,\n * so a manifest without it means \"not published\", never \"empty\": consumers\n * must not treat its absence as zero rows, or as rows missing locally.\n * The core datasets are always published; a manifest that lacks one of them\n * is incomplete, and consumers say so.\n */\nconst OPTIONAL_DATASET: Record<ArchiveDataset, boolean> = {\n\tblocks: false,\n\ttransactions: false,\n\tevents: false,\n\tstate_writes: true,\n};\n\nexport function isOptionalDataset(dataset: ArchiveDataset): boolean {\n\treturn OPTIONAL_DATASET[dataset];\n}\n\n/**\n * Whether a manifest publishes a dataset: always for a core dataset, and for\n * an optional one only when it names at least one partition or zero-record\n * range of it.\n */\nexport function manifestPublishes(\n\tmanifest: {\n\t\tpartitions?: ReadonlyArray<{ dataset: string }>;\n\t\tzero_record_ranges?: ReadonlyArray<{ dataset: string }>;\n\t},\n\tdataset: ArchiveDataset,\n): boolean {\n\tif (!isOptionalDataset(dataset)) return true;\n\treturn (\n\t\t(manifest.partitions ?? []).some((p) => p.dataset === dataset) ||\n\t\t(manifest.zero_record_ranges ?? []).some((z) => z.dataset === dataset)\n\t);\n}\n\n/**\n * The column a dataset's rows are placed at a height by. Each dataset is\n * stored in the table of the same name.\n */\nexport function archiveHeightColumn(\n\tdataset: ArchiveDataset,\n): \"height\" | \"block_height\" {\n\tswitch (dataset) {\n\t\tcase \"blocks\":\n\t\t\treturn \"height\";\n\t\tcase \"transactions\":\n\t\tcase \"events\":\n\t\tcase \"state_writes\":\n\t\t\treturn \"block_height\";\n\t\tdefault:\n\t\t\treturn assertNeverDataset(dataset);\n\t}\n}\n\nexport function assertNeverDataset(dataset: never): never {\n\tthrow new Error(`unhandled archive dataset: ${String(dataset)}`);\n}\n"
   ],
-  "mappings": ";AAAA;AA8BO,SAAS,QAAQ,CAAC,OAAmD;AAAA,EAC3E,IAAI,UAAU,QAAQ,UAAU;AAAA,IAAW,OAAO;AAAA,EAClD,MAAM,OAAO,OAAO,UAAU,WAAW,OAAO,KAAK,IAAI;AAAA,EACzD,OAAO,IAAI,KAAK,QAAQ,MAAM,IAAI;AAAA;AAGnC,SAAS,QAAQ,CAAC,OAAwB;AAAA,EACzC,OAAO,OAAO,UAAU,WAAW,OAAO,KAAK,IAAI,OAAO,KAAe;AAAA;AAG1E,SAAS,MAAM,CAAC,OAAwB;AAAA,EACvC,OAAO,OAAO,SAAS,KAAK,IAAI,MAAM,SAAS,MAAM,IAAI,OAAO,KAAK;AAAA;AAGtE,SAAS,cAAc,CAAC,OAA+B;AAAA,EACtD,OAAO,UAAU,QAAQ,UAAU,YAAY,OAAO,OAAO,KAAK;AAAA;AAG5D,SAAS,SAAS,CAAC,SAAyB,KAAyB;AAAA,EAC3E,IAAI,YAAY,UAAU;AAAA,IACzB,OAAO;AAAA,MACN,SAAS,SAAS,IAAI,MAAM,CAAC;AAAA,MAC7B,SAAS,OAAO,IAAI,IAAI,CAAC;AAAA,MACzB,SAAS,OAAO,IAAI,WAAW,CAAC;AAAA,MAChC,SAAS,SAAS,IAAI,iBAAiB,CAAC;AAAA,MACxC,SAAS,eAAe,IAAI,eAAe,CAAC;AAAA,MAC5C,SAAS,eAAe,IAAI,gBAAgB,CAAC;AAAA,MAC7C,SAAS,SAAS,IAAI,SAAS,CAAC;AAAA,IACjC,EAAE,KAAK,GAAG;AAAA,EACX;AAAA,EACA,IAAI,YAAY,gBAAgB;AAAA,IAC/B,OAAO;AAAA,MACN,SAAS,OAAO,IAAI,KAAK,CAAC;AAAA,MAC1B,SAAS,SAAS,IAAI,YAAY,CAAC;AAAA,MACnC,SAAS,SAAS,IAAI,QAAQ,CAAC;AAAA,MAC/B,SAAS,OAAO,IAAI,IAAI,CAAC;AAAA,MACzB,SAAS,OAAO,IAAI,MAAM,CAAC;AAAA,MAC3B,SAAS,OAAO,IAAI,MAAM,CAAC;AAAA,MAC3B,SAAS,eAAe,IAAI,WAAW,CAAC;AAAA,MACxC,SAAS,eAAe,IAAI,aAAa,CAAC;AAAA,MAC1C,SAAS,eAAe,IAAI,kBAAkB,CAAC;AAAA,MAC/C,SAAS,eAAe,IAAI,UAAU,CAAC;AAAA,MACvC,SAAS,OAAO,IAAI,MAAM,CAAC;AAAA,IAC5B,EAAE,KAAK,GAAG;AAAA,EACX;AAAA,EACA,OAAO;AAAA,IACN,SAAS,OAAO,IAAI,KAAK,CAAC;AAAA,IAC1B,SAAS,SAAS,IAAI,YAAY,CAAC;AAAA,IACnC,SAAS,SAAS,IAAI,WAAW,CAAC;AAAA,IAClC,SAAS,OAAO,IAAI,UAAU,CAAC;AAAA,IAC/B,SAAS,OAAO,IAAI,SAAS,CAAC;AAAA,EAC/B,EAAE,KAAK,GAAG;AAAA;AAQJ,SAAS,aAAa,CAAC,SAAiC;AAAA,EAC9D,IAAI,YAAY,UAAU;AAAA,IACzB,OAAO;AAAA,EACR;AAAA,EACA,IAAI,YAAY,gBAAgB;AAAA,IAC/B,OAAO;AAAA,EACR;AAAA,EACA,OAAO;AAAA;AAYR,eAAsB,qBAAqB,CAAC,QAMxB;AAAA,EACnB,QAAQ,UAAU,SAAS,SAAS;AAAA,EACpC,MAAM,gBAAgB,OAAO,iBAAiB;AAAA,EAC9C,IAAI,QAAQ;AAAA,EACZ,IAAI;AAAA,IACH,iBAAiB,OAAO,MAAM;AAAA,MAC7B,MAAM,OAAO,GAAG,UAAU,SAAS,GAAG;AAAA;AAAA,MACtC,IAAI,CAAC,SAAS,MAAM,IAAI,GAAG;AAAA,QAC1B,MAAM,KAAK,UAAU,OAAO;AAAA,MAC7B;AAAA,MACA;AAAA,MACA,IAAI,OAAO,cAAc,QAAQ,kBAAkB,GAAG;AAAA,QACrD,OAAO,WAAW,KAAK;AAAA,MACxB;AAAA,IACD;AAAA,IACA,MAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAAA,MAC5C,SAAS,IAAI,CAAC,UACb,QAAQ,OAAO,KAAK,IAAI,QAAQ,CACjC;AAAA,KACA;AAAA,IACA,OAAO,OAAO;AAAA,IACf,SAAS,QAAQ,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC,CAAC;AAAA,IAC1E,MAAM;AAAA;AAAA,EAEP,OAAO;AAAA;",
-  "debugId": "4643C93B767098ED64756E2164756E21",
+  "mappings": ";AAAA;;;ACYO,IAAM,mBAAmB;AAAA,EAC/B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACD;AAKO,SAAS,gBAAgB,CAAC,OAAwC;AAAA,EACxE,OAAQ,iBAAuC,SAAS,KAAK;AAAA;AAW9D,IAAM,mBAAoD;AAAA,EACzD,QAAQ;AAAA,EACR,cAAc;AAAA,EACd,QAAQ;AAAA,EACR,cAAc;AACf;AAEO,SAAS,iBAAiB,CAAC,SAAkC;AAAA,EACnE,OAAO,iBAAiB;AAAA;AAQlB,SAAS,iBAAiB,CAChC,UAIA,SACU;AAAA,EACV,IAAI,CAAC,kBAAkB,OAAO;AAAA,IAAG,OAAO;AAAA,EACxC,QACE,SAAS,cAAc,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,YAAY,OAAO,MAC5D,SAAS,sBAAsB,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,YAAY,OAAO;AAAA;AAQhE,SAAS,mBAAmB,CAClC,SAC4B;AAAA,EAC5B,QAAQ;AAAA,SACF;AAAA,MACJ,OAAO;AAAA,SACH;AAAA,SACA;AAAA,SACA;AAAA,MACJ,OAAO;AAAA;AAAA,MAEP,OAAO,mBAAmB,OAAO;AAAA;AAAA;AAI7B,SAAS,kBAAkB,CAAC,SAAuB;AAAA,EACzD,MAAM,IAAI,MAAM,8BAA8B,OAAO,OAAO,GAAG;AAAA;;;ADtDzD,SAAS,QAAQ,CAAC,OAAmD;AAAA,EAC3E,IAAI,UAAU,QAAQ,UAAU;AAAA,IAAW,OAAO;AAAA,EAClD,MAAM,OAAO,OAAO,UAAU,WAAW,OAAO,KAAK,IAAI;AAAA,EACzD,OAAO,IAAI,KAAK,QAAQ,MAAM,IAAI;AAAA;AAGnC,SAAS,QAAQ,CAAC,OAAwB;AAAA,EACzC,OAAO,OAAO,UAAU,WAAW,OAAO,KAAK,IAAI,OAAO,KAAe;AAAA;AAG1E,SAAS,gBAAgB,CAAC,OAA+B;AAAA,EACxD,OAAO,UAAU,QAAQ,UAAU,YAAY,OAAO,SAAS,KAAK;AAAA;AAGrE,SAAS,MAAM,CAAC,OAAwB;AAAA,EACvC,OAAO,OAAO,SAAS,KAAK,IAAI,MAAM,SAAS,MAAM,IAAI,OAAO,KAAK;AAAA;AAGtE,SAAS,cAAc,CAAC,OAA+B;AAAA,EACtD,OAAO,UAAU,QAAQ,UAAU,YAAY,OAAO,OAAO,KAAK;AAAA;AAG5D,SAAS,SAAS,CAAC,SAAyB,KAAyB;AAAA,EAC3E,QAAQ;AAAA,SACF;AAAA,MACJ,OAAO;AAAA,QACN,SAAS,SAAS,IAAI,MAAM,CAAC;AAAA,QAC7B,SAAS,OAAO,IAAI,IAAI,CAAC;AAAA,QACzB,SAAS,OAAO,IAAI,WAAW,CAAC;AAAA,QAChC,SAAS,SAAS,IAAI,iBAAiB,CAAC;AAAA,QACxC,SAAS,eAAe,IAAI,eAAe,CAAC;AAAA,QAC5C,SAAS,eAAe,IAAI,gBAAgB,CAAC;AAAA,QAC7C,SAAS,SAAS,IAAI,SAAS,CAAC;AAAA,MACjC,EAAE,KAAK,GAAG;AAAA,SACN;AAAA,MACJ,OAAO;AAAA,QACN,SAAS,OAAO,IAAI,KAAK,CAAC;AAAA,QAC1B,SAAS,SAAS,IAAI,YAAY,CAAC;AAAA,QACnC,SAAS,SAAS,IAAI,QAAQ,CAAC;AAAA,QAC/B,SAAS,OAAO,IAAI,IAAI,CAAC;AAAA,QACzB,SAAS,OAAO,IAAI,MAAM,CAAC;AAAA,QAC3B,SAAS,OAAO,IAAI,MAAM,CAAC;AAAA,QAC3B,SAAS,eAAe,IAAI,WAAW,CAAC;AAAA,QACxC,SAAS,eAAe,IAAI,aAAa,CAAC;AAAA,QAC1C,SAAS,eAAe,IAAI,kBAAkB,CAAC;AAAA,QAC/C,SAAS,eAAe,IAAI,UAAU,CAAC;AAAA,QACvC,SAAS,OAAO,IAAI,MAAM,CAAC;AAAA,MAC5B,EAAE,KAAK,GAAG;AAAA,SACN;AAAA,MACJ,OAAO;AAAA,QACN,SAAS,OAAO,IAAI,KAAK,CAAC;AAAA,QAC1B,SAAS,SAAS,IAAI,YAAY,CAAC;AAAA,QACnC,SAAS,SAAS,IAAI,WAAW,CAAC;AAAA,QAClC,SAAS,OAAO,IAAI,UAAU,CAAC;AAAA,QAC/B,SAAS,OAAO,IAAI,SAAS,CAAC;AAAA,MAC/B,EAAE,KAAK,GAAG;AAAA,SACN;AAAA,MACJ,OAAO;AAAA,QACN,SAAS,SAAS,IAAI,YAAY,CAAC;AAAA,QACnC,SAAS,SAAS,IAAI,OAAO,CAAC;AAAA,QAC9B,SAAS,iBAAiB,IAAI,QAAQ,CAAC;AAAA,QACvC,SAAS,OAAO,IAAI,GAAG,CAAC;AAAA,QACxB,SAAS,OAAO,IAAI,SAAS,CAAC;AAAA,MAC/B,EAAE,KAAK,GAAG;AAAA;AAAA,MAEV,OAAO,mBAAmB,OAAO;AAAA;AAAA;AAS7B,SAAS,aAAa,CAAC,SAAiC;AAAA,EAC9D,QAAQ;AAAA,SACF;AAAA,MACJ,OAAO;AAAA,SACH;AAAA,MACJ,OAAO;AAAA,SACH;AAAA,MACJ,OAAO;AAAA,SACH;AAAA,MACJ,OAAO;AAAA;AAAA,MAEP,OAAO,mBAAmB,OAAO;AAAA;AAAA;AAapC,eAAsB,qBAAqB,CAAC,QAMxB;AAAA,EACnB,QAAQ,UAAU,SAAS,SAAS;AAAA,EACpC,MAAM,gBAAgB,OAAO,iBAAiB;AAAA,EAC9C,IAAI,QAAQ;AAAA,EACZ,IAAI;AAAA,IACH,iBAAiB,OAAO,MAAM;AAAA,MAC7B,MAAM,OAAO,GAAG,UAAU,SAAS,GAAG;AAAA;AAAA,MACtC,IAAI,CAAC,SAAS,MAAM,IAAI,GAAG;AAAA,QAC1B,MAAM,KAAK,UAAU,OAAO;AAAA,MAC7B;AAAA,MACA;AAAA,MACA,IAAI,OAAO,cAAc,QAAQ,kBAAkB,GAAG;AAAA,QACrD,OAAO,WAAW,KAAK;AAAA,MACxB;AAAA,IACD;AAAA,IACA,MAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAAA,MAC5C,SAAS,IAAI,CAAC,UACb,QAAQ,OAAO,KAAK,IAAI,QAAQ,CACjC;AAAA,KACA;AAAA,IACA,OAAO,OAAO;AAAA,IACf,SAAS,QAAQ,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC,CAAC;AAAA,IAC1E,MAAM;AAAA;AAAA,EAEP,OAAO;AAAA;",
+  "debugId": "63CC074406EEB2EC64756E2164756E21",
   "names": []
 }

package/dist/src/archive/datasets.d.ts

@@ -0,0 +1,37 @@
+/**
+* The canonical archive's datasets — the one list every exporter, loader,
+* verifier and fetch gate derives from.
+*
+* Order is load order: each dataset's foreign keys point only at datasets
+* earlier in the list, so loading in this order never violates a constraint.
+*
+* Per-dataset code branches with an exhaustive `switch` ending in
+* `assertNeverDataset`, never a trailing default: adding a dataset here must
+* fail to compile everywhere it needs its own schema, SQL or table, instead of
+* silently inheriting another dataset's.
+*/
+declare const ARCHIVE_DATASETS: readonly ["blocks", "transactions", "events", "state_writes"];
+/** One archive dataset: blocks, transactions, events or state writes. */
+type ArchiveDataset = (typeof ARCHIVE_DATASETS)[number];
+declare function isArchiveDataset(value: string): value is ArchiveDataset;
+declare function isOptionalDataset(dataset: ArchiveDataset): boolean;
+/**
+* Whether a manifest publishes a dataset: always for a core dataset, and for
+* an optional one only when it names at least one partition or zero-record
+* range of it.
+*/
+declare function manifestPublishes(manifest: {
+	partitions?: ReadonlyArray<{
+		dataset: string
+	}>
+	zero_record_ranges?: ReadonlyArray<{
+		dataset: string
+	}>
+}, dataset: ArchiveDataset): boolean;
+/**
+* The column a dataset's rows are placed at a height by. Each dataset is
+* stored in the table of the same name.
+*/
+declare function archiveHeightColumn(dataset: ArchiveDataset): "height" | "block_height";
+declare function assertNeverDataset(dataset: never): never;
+export { ARCHIVE_DATASETS, ArchiveDataset, archiveHeightColumn, assertNeverDataset, isArchiveDataset, isOptionalDataset, manifestPublishes };

package/dist/src/archive/datasets.js

@@ -0,0 +1,50 @@
+// src/archive/datasets.ts
+var ARCHIVE_DATASETS = [
+  "blocks",
+  "transactions",
+  "events",
+  "state_writes"
+];
+function isArchiveDataset(value) {
+  return ARCHIVE_DATASETS.includes(value);
+}
+var OPTIONAL_DATASET = {
+  blocks: false,
+  transactions: false,
+  events: false,
+  state_writes: true
+};
+function isOptionalDataset(dataset) {
+  return OPTIONAL_DATASET[dataset];
+}
+function manifestPublishes(manifest, dataset) {
+  if (!isOptionalDataset(dataset))
+    return true;
+  return (manifest.partitions ?? []).some((p) => p.dataset === dataset) || (manifest.zero_record_ranges ?? []).some((z) => z.dataset === dataset);
+}
+function archiveHeightColumn(dataset) {
+  switch (dataset) {
+    case "blocks":
+      return "height";
+    case "transactions":
+    case "events":
+    case "state_writes":
+      return "block_height";
+    default:
+      return assertNeverDataset(dataset);
+  }
+}
+function assertNeverDataset(dataset) {
+  throw new Error(`unhandled archive dataset: ${String(dataset)}`);
+}
+export {
+  ARCHIVE_DATASETS,
+  archiveHeightColumn,
+  assertNeverDataset,
+  isArchiveDataset,
+  isOptionalDataset,
+  manifestPublishes
+};
+
+//# debugId=EF9D77F8C69ED8D264756E2164756E21
+//# sourceMappingURL=datasets.js.map

package/dist/src/archive/datasets.js.map

@@ -0,0 +1,10 @@
+{
+  "version": 3,
+  "sources": ["../src/archive/datasets.ts"],
+  "sourcesContent": [
+    "/**\n * The canonical archive's datasets — the one list every exporter, loader,\n * verifier and fetch gate derives from.\n *\n * Order is load order: each dataset's foreign keys point only at datasets\n * earlier in the list, so loading in this order never violates a constraint.\n *\n * Per-dataset code branches with an exhaustive `switch` ending in\n * `assertNeverDataset`, never a trailing default: adding a dataset here must\n * fail to compile everywhere it needs its own schema, SQL or table, instead of\n * silently inheriting another dataset's.\n */\nexport const ARCHIVE_DATASETS = [\n\t\"blocks\",\n\t\"transactions\",\n\t\"events\",\n\t\"state_writes\",\n] as const;\n\n/** One archive dataset: blocks, transactions, events or state writes. */\nexport type ArchiveDataset = (typeof ARCHIVE_DATASETS)[number];\n\nexport function isArchiveDataset(value: string): value is ArchiveDataset {\n\treturn (ARCHIVE_DATASETS as readonly string[]).includes(value);\n}\n\n/**\n * Datasets a manifest may leave out. The exporter publishes `state_writes`\n * only from an instance that holds it from the start of the exported range,\n * so a manifest without it means \"not published\", never \"empty\": consumers\n * must not treat its absence as zero rows, or as rows missing locally.\n * The core datasets are always published; a manifest that lacks one of them\n * is incomplete, and consumers say so.\n */\nconst OPTIONAL_DATASET: Record<ArchiveDataset, boolean> = {\n\tblocks: false,\n\ttransactions: false,\n\tevents: false,\n\tstate_writes: true,\n};\n\nexport function isOptionalDataset(dataset: ArchiveDataset): boolean {\n\treturn OPTIONAL_DATASET[dataset];\n}\n\n/**\n * Whether a manifest publishes a dataset: always for a core dataset, and for\n * an optional one only when it names at least one partition or zero-record\n * range of it.\n */\nexport function manifestPublishes(\n\tmanifest: {\n\t\tpartitions?: ReadonlyArray<{ dataset: string }>;\n\t\tzero_record_ranges?: ReadonlyArray<{ dataset: string }>;\n\t},\n\tdataset: ArchiveDataset,\n): boolean {\n\tif (!isOptionalDataset(dataset)) return true;\n\treturn (\n\t\t(manifest.partitions ?? []).some((p) => p.dataset === dataset) ||\n\t\t(manifest.zero_record_ranges ?? []).some((z) => z.dataset === dataset)\n\t);\n}\n\n/**\n * The column a dataset's rows are placed at a height by. Each dataset is\n * stored in the table of the same name.\n */\nexport function archiveHeightColumn(\n\tdataset: ArchiveDataset,\n): \"height\" | \"block_height\" {\n\tswitch (dataset) {\n\t\tcase \"blocks\":\n\t\t\treturn \"height\";\n\t\tcase \"transactions\":\n\t\tcase \"events\":\n\t\tcase \"state_writes\":\n\t\t\treturn \"block_height\";\n\t\tdefault:\n\t\t\treturn assertNeverDataset(dataset);\n\t}\n}\n\nexport function assertNeverDataset(dataset: never): never {\n\tthrow new Error(`unhandled archive dataset: ${String(dataset)}`);\n}\n"
+  ],
+  "mappings": ";AAYO,IAAM,mBAAmB;AAAA,EAC/B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACD;AAKO,SAAS,gBAAgB,CAAC,OAAwC;AAAA,EACxE,OAAQ,iBAAuC,SAAS,KAAK;AAAA;AAW9D,IAAM,mBAAoD;AAAA,EACzD,QAAQ;AAAA,EACR,cAAc;AAAA,EACd,QAAQ;AAAA,EACR,cAAc;AACf;AAEO,SAAS,iBAAiB,CAAC,SAAkC;AAAA,EACnE,OAAO,iBAAiB;AAAA;AAQlB,SAAS,iBAAiB,CAChC,UAIA,SACU;AAAA,EACV,IAAI,CAAC,kBAAkB,OAAO;AAAA,IAAG,OAAO;AAAA,EACxC,QACE,SAAS,cAAc,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,YAAY,OAAO,MAC5D,SAAS,sBAAsB,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,YAAY,OAAO;AAAA;AAQhE,SAAS,mBAAmB,CAClC,SAC4B;AAAA,EAC5B,QAAQ;AAAA,SACF;AAAA,MACJ,OAAO;AAAA,SACH;AAAA,SACA;AAAA,SACA;AAAA,MACJ,OAAO;AAAA;AAAA,MAEP,OAAO,mBAAmB,OAAO;AAAA;AAAA;AAI7B,SAAS,kBAAkB,CAAC,SAAuB;AAAA,EACzD,MAAM,IAAI,MAAM,8BAA8B,OAAO,OAAO,GAAG;AAAA;",
+  "debugId": "EF9D77F8C69ED8D264756E2164756E21",
+  "names": []
+}

package/dist/src/archive/derived-stage-report.d.ts

@@ -1,35 +1,21 @@
 /**
-* Derived-stage acceptance — what each service reads, writes, and needs
-* beyond the canonical archive.
+* The canonical archive's datasets — the one list every exporter, loader,
+* verifier and fetch gate derives from.
 *
-* The canonical archive rebuilds `blocks`, `transactions`, and `events`
-* deterministically from R2 alone. Every stage downstream of those tables
-* (decoders, subgraphs, webhook delivery) has its own inputs, outputs,
-* and rebuild requirements — and an operator restoring from the archive
-* needs to know exactly which of those it also needs to source, replay, or
-* re-derive.
+* Order is load order: each dataset's foreign keys point only at datasets
+* earlier in the list, so loading in this order never violates a constraint.
 *
-* This module defines the reporting shape. Each producer/service declares
-* itself once (a `ServiceManifest`) and the aggregator emits one report per
-* service to `reports/derived/<service>.json`. Stateful verification —
-* actually re-running the service and comparing outputs — is Phase 4 work
-* (see P4.10 effect manifests, P4.13 stateful deep verify); this shape is
-* what Phase 4 will populate.
-*
-* The report answers three operator questions:
-*   1. What canonical inputs does this service consume? (archive suffices)
-*   2. What external inputs does it need? (RPC endpoints, envelope journal,
-*      contract source, ORACLE feeds — anything not in the archive)
-*   3. If I restore from R2 alone, can this service resume? If not, what
-*      is missing?
-*
-* Silence is the failure mode we most want to prevent: a service that ships
-* without a manifest is a service whose rebuild story is undocumented.
-* `assertAllServicesDeclared()` fails a build when a known service is missing.
+* Per-dataset code branches with an exhaustive `switch` ending in
+* `assertNeverDataset`, never a trailing default: adding a dataset here must
+* fail to compile everywhere it needs its own schema, SQL or table, instead of
+* silently inheriting another dataset's.
 */
+declare const ARCHIVE_DATASETS: readonly ["blocks", "transactions", "events", "state_writes"];
+/** One archive dataset: blocks, transactions, events or state writes. */
+type ArchiveDataset = (typeof ARCHIVE_DATASETS)[number];
 declare const DERIVED_STAGE_REPORT_SCHEMA_VERSION: 1;
 type ServiceKind = "decoder" | "subgraph-runtime" | "webhook-runtime" | "protocol-producer";
-type CanonicalInput = "blocks" | "transactions" | "events";
+type CanonicalInput = ArchiveDataset;
 interface ExternalInput {
 	/** Short name for the dependency: `stacks-node-rpc`, `hiro-extended-api`,
 	*  `observer-journal`, `contract-source-registry`, etc. */

package/dist/src/archive/derived-stage-report.js

@@ -1,3 +1,43 @@
+// src/archive/datasets.ts
+var ARCHIVE_DATASETS = [
+  "blocks",
+  "transactions",
+  "events",
+  "state_writes"
+];
+function isArchiveDataset(value) {
+  return ARCHIVE_DATASETS.includes(value);
+}
+var OPTIONAL_DATASET = {
+  blocks: false,
+  transactions: false,
+  events: false,
+  state_writes: true
+};
+function isOptionalDataset(dataset) {
+  return OPTIONAL_DATASET[dataset];
+}
+function manifestPublishes(manifest, dataset) {
+  if (!isOptionalDataset(dataset))
+    return true;
+  return (manifest.partitions ?? []).some((p) => p.dataset === dataset) || (manifest.zero_record_ranges ?? []).some((z) => z.dataset === dataset);
+}
+function archiveHeightColumn(dataset) {
+  switch (dataset) {
+    case "blocks":
+      return "height";
+    case "transactions":
+    case "events":
+    case "state_writes":
+      return "block_height";
+    default:
+      return assertNeverDataset(dataset);
+  }
+}
+function assertNeverDataset(dataset) {
+  throw new Error(`unhandled archive dataset: ${String(dataset)}`);
+}
+
 // src/archive/derived-stage-report.ts
 var DERIVED_STAGE_REPORT_SCHEMA_VERSION = 1;
 function buildDerivedStageReport(manifests, generatedAt) {
@@ -11,11 +51,7 @@ function buildDerivedStageReport(manifests, generatedAt) {
 function validateServiceManifests(manifests) {
   const issues = [];
   const seen = new Set;
-  const canonicalDatasets = new Set([
-    "blocks",
-    "transactions",
-    "events"
-  ]);
+  const canonicalDatasets = new Set(ARCHIVE_DATASETS);
   for (const m of manifests) {
     if (seen.has(m.name)) {
       issues.push({ name: m.name, problem: "duplicate service name" });
@@ -44,5 +80,5 @@ export {
   validateServiceManifests
 };
 
-//# debugId=664A04684A589A7B64756E2164756E21
+//# debugId=4E289EF15D65D38E64756E2164756E21
 //# sourceMappingURL=derived-stage-report.js.map

package/dist/src/archive/derived-stage-report.js.map

@@ -1,10 +1,11 @@
 {
   "version": 3,
-  "sources": ["../src/archive/derived-stage-report.ts"],
+  "sources": ["../src/archive/datasets.ts", "../src/archive/derived-stage-report.ts"],
   "sourcesContent": [
-    "/**\n * Derived-stage acceptance — what each service reads, writes, and needs\n * beyond the canonical archive.\n *\n * The canonical archive rebuilds `blocks`, `transactions`, and `events`\n * deterministically from R2 alone. Every stage downstream of those tables\n * (decoders, subgraphs, webhook delivery) has its own inputs, outputs,\n * and rebuild requirements — and an operator restoring from the archive\n * needs to know exactly which of those it also needs to source, replay, or\n * re-derive.\n *\n * This module defines the reporting shape. Each producer/service declares\n * itself once (a `ServiceManifest`) and the aggregator emits one report per\n * service to `reports/derived/<service>.json`. Stateful verification —\n * actually re-running the service and comparing outputs — is Phase 4 work\n * (see P4.10 effect manifests, P4.13 stateful deep verify); this shape is\n * what Phase 4 will populate.\n *\n * The report answers three operator questions:\n *   1. What canonical inputs does this service consume? (archive suffices)\n *   2. What external inputs does it need? (RPC endpoints, envelope journal,\n *      contract source, ORACLE feeds — anything not in the archive)\n *   3. If I restore from R2 alone, can this service resume? If not, what\n *      is missing?\n *\n * Silence is the failure mode we most want to prevent: a service that ships\n * without a manifest is a service whose rebuild story is undocumented.\n * `assertAllServicesDeclared()` fails a build when a known service is missing.\n */\n\nexport const DERIVED_STAGE_REPORT_SCHEMA_VERSION = 1 as const;\n\nexport type ServiceKind =\n\t| \"decoder\"\n\t| \"subgraph-runtime\"\n\t| \"webhook-runtime\"\n\t| \"protocol-producer\";\n\nexport type CanonicalInput = \"blocks\" | \"transactions\" | \"events\";\n\nexport interface ExternalInput {\n\t/** Short name for the dependency: `stacks-node-rpc`, `hiro-extended-api`,\n\t *  `observer-journal`, `contract-source-registry`, etc. */\n\tname: string;\n\t/** Why the archive can't provide this. One sentence. */\n\treason: string;\n\t/** Where an operator can source it. `bundled-stacks`, `own-node`,\n\t *  `hiro-public-api`, `re-derive-from-events`, `manual-import`. */\n\tsource: string;\n\t/** True if the archive supplies enough to re-derive this dependency\n\t *  without a live external source. Rare but happens (e.g. contract ABIs\n\t *  derivable from `transactions.raw_tx`). */\n\trebuildable_from_archive: boolean;\n}\n\nexport interface Output {\n\t/** Table/topic/queue name in the operator's runtime. */\n\ttarget: string;\n\t/** How the operator verifies this output — e.g. `row-count`,\n\t *  `semantic-digest`, `effect-manifest`, `dlq-empty`. */\n\tverification: string;\n}\n\nexport interface ServiceManifest {\n\tname: string;\n\tkind: ServiceKind;\n\tdescription: string;\n\tcanonical_inputs: CanonicalInput[];\n\texternal_inputs: ExternalInput[];\n\toutputs: Output[];\n\t/**\n\t * Explicit judgment on whether R2 alone can rebuild this service.\n\t * `false` REQUIRES `external_inputs` to name what else is needed.\n\t */\n\tr2_alone_can_rebuild: boolean;\n\t/** When true, this service can only replay forward from a starting point\n\t *  the operator supplies (e.g. a webhook consumer's last-delivered id).\n\t *  Publish it, don't hide it. */\n\trequires_operator_state: boolean;\n\t/** Repo path where the service lives, for cross-reference. */\n\tsource_path: string;\n}\n\nexport interface DerivedStageReport {\n\tschema_version: typeof DERIVED_STAGE_REPORT_SCHEMA_VERSION;\n\tgenerated_at: string;\n\tservices: ServiceManifest[];\n}\n\n/**\n * Compose a report from a set of registered manifests. Sorts services by name\n * so the JSON is stable across runs — a re-emit with unchanged manifests\n * produces byte-identical output, so an operator can diff two archive\n * publish cycles cleanly.\n */\nexport function buildDerivedStageReport(\n\tmanifests: ReadonlyArray<ServiceManifest>,\n\tgeneratedAt: string,\n): DerivedStageReport {\n\tconst sorted = [...manifests].sort((a, b) => a.name.localeCompare(b.name));\n\treturn {\n\t\tschema_version: DERIVED_STAGE_REPORT_SCHEMA_VERSION,\n\t\tgenerated_at: generatedAt,\n\t\tservices: sorted,\n\t};\n}\n\nexport type ManifestValidationIssue = {\n\tname: string;\n\tproblem: string;\n};\n\n/**\n * Enforce the invariants the report shape leaves ambiguous:\n *   - `r2_alone_can_rebuild: false` REQUIRES at least one external input.\n *     Otherwise the report claims a rebuild gap it can't explain, which is\n *     worse than declaring nothing.\n *   - Names are unique — a duplicate is almost always a copy-paste that\n *     silently shadows the earlier declaration.\n *   - Every canonical_input is a real dataset name.\n */\nexport function validateServiceManifests(\n\tmanifests: ReadonlyArray<ServiceManifest>,\n): ManifestValidationIssue[] {\n\tconst issues: ManifestValidationIssue[] = [];\n\tconst seen = new Set<string>();\n\tconst canonicalDatasets: ReadonlySet<CanonicalInput> = new Set([\n\t\t\"blocks\",\n\t\t\"transactions\",\n\t\t\"events\",\n\t]);\n\tfor (const m of manifests) {\n\t\tif (seen.has(m.name)) {\n\t\t\tissues.push({ name: m.name, problem: \"duplicate service name\" });\n\t\t}\n\t\tseen.add(m.name);\n\t\tif (!m.r2_alone_can_rebuild && m.external_inputs.length === 0) {\n\t\t\tissues.push({\n\t\t\t\tname: m.name,\n\t\t\t\tproblem:\n\t\t\t\t\t\"declares r2_alone_can_rebuild=false but lists no external_inputs — the gap is undocumented\",\n\t\t\t});\n\t\t}\n\t\tfor (const input of m.canonical_inputs) {\n\t\t\tif (!canonicalDatasets.has(input)) {\n\t\t\t\tissues.push({\n\t\t\t\t\tname: m.name,\n\t\t\t\t\tproblem: `canonical_inputs contains unknown dataset: ${String(input)}`,\n\t\t\t\t});\n\t\t\t}\n\t\t}\n\t}\n\treturn issues;\n}\n"
+    "/**\n * The canonical archive's datasets — the one list every exporter, loader,\n * verifier and fetch gate derives from.\n *\n * Order is load order: each dataset's foreign keys point only at datasets\n * earlier in the list, so loading in this order never violates a constraint.\n *\n * Per-dataset code branches with an exhaustive `switch` ending in\n * `assertNeverDataset`, never a trailing default: adding a dataset here must\n * fail to compile everywhere it needs its own schema, SQL or table, instead of\n * silently inheriting another dataset's.\n */\nexport const ARCHIVE_DATASETS = [\n\t\"blocks\",\n\t\"transactions\",\n\t\"events\",\n\t\"state_writes\",\n] as const;\n\n/** One archive dataset: blocks, transactions, events or state writes. */\nexport type ArchiveDataset = (typeof ARCHIVE_DATASETS)[number];\n\nexport function isArchiveDataset(value: string): value is ArchiveDataset {\n\treturn (ARCHIVE_DATASETS as readonly string[]).includes(value);\n}\n\n/**\n * Datasets a manifest may leave out. The exporter publishes `state_writes`\n * only from an instance that holds it from the start of the exported range,\n * so a manifest without it means \"not published\", never \"empty\": consumers\n * must not treat its absence as zero rows, or as rows missing locally.\n * The core datasets are always published; a manifest that lacks one of them\n * is incomplete, and consumers say so.\n */\nconst OPTIONAL_DATASET: Record<ArchiveDataset, boolean> = {\n\tblocks: false,\n\ttransactions: false,\n\tevents: false,\n\tstate_writes: true,\n};\n\nexport function isOptionalDataset(dataset: ArchiveDataset): boolean {\n\treturn OPTIONAL_DATASET[dataset];\n}\n\n/**\n * Whether a manifest publishes a dataset: always for a core dataset, and for\n * an optional one only when it names at least one partition or zero-record\n * range of it.\n */\nexport function manifestPublishes(\n\tmanifest: {\n\t\tpartitions?: ReadonlyArray<{ dataset: string }>;\n\t\tzero_record_ranges?: ReadonlyArray<{ dataset: string }>;\n\t},\n\tdataset: ArchiveDataset,\n): boolean {\n\tif (!isOptionalDataset(dataset)) return true;\n\treturn (\n\t\t(manifest.partitions ?? []).some((p) => p.dataset === dataset) ||\n\t\t(manifest.zero_record_ranges ?? []).some((z) => z.dataset === dataset)\n\t);\n}\n\n/**\n * The column a dataset's rows are placed at a height by. Each dataset is\n * stored in the table of the same name.\n */\nexport function archiveHeightColumn(\n\tdataset: ArchiveDataset,\n): \"height\" | \"block_height\" {\n\tswitch (dataset) {\n\t\tcase \"blocks\":\n\t\t\treturn \"height\";\n\t\tcase \"transactions\":\n\t\tcase \"events\":\n\t\tcase \"state_writes\":\n\t\t\treturn \"block_height\";\n\t\tdefault:\n\t\t\treturn assertNeverDataset(dataset);\n\t}\n}\n\nexport function assertNeverDataset(dataset: never): never {\n\tthrow new Error(`unhandled archive dataset: ${String(dataset)}`);\n}\n",
+    "/**\n * Derived-stage acceptance — what each service reads, writes, and needs\n * beyond the canonical archive.\n *\n * The canonical archive rebuilds `blocks`, `transactions`, and `events`\n * deterministically from R2 alone. Every stage downstream of those tables\n * (decoders, subgraphs, webhook delivery) has its own inputs, outputs,\n * and rebuild requirements — and an operator restoring from the archive\n * needs to know exactly which of those it also needs to source, replay, or\n * re-derive.\n *\n * This module defines the reporting shape. Each producer/service declares\n * itself once (a `ServiceManifest`) and the aggregator emits one report per\n * service to `reports/derived/<service>.json`. Stateful verification —\n * actually re-running the service and comparing outputs — is Phase 4 work\n * (see P4.10 effect manifests, P4.13 stateful deep verify); this shape is\n * what Phase 4 will populate.\n *\n * The report answers three operator questions:\n *   1. What canonical inputs does this service consume? (archive suffices)\n *   2. What external inputs does it need? (RPC endpoints, envelope journal,\n *      contract source, ORACLE feeds — anything not in the archive)\n *   3. If I restore from R2 alone, can this service resume? If not, what\n *      is missing?\n *\n * Silence is the failure mode we most want to prevent: a service that ships\n * without a manifest is a service whose rebuild story is undocumented.\n * `assertAllServicesDeclared()` fails a build when a known service is missing.\n */\n\nimport { ARCHIVE_DATASETS, type ArchiveDataset } from \"./datasets.ts\";\n\nexport const DERIVED_STAGE_REPORT_SCHEMA_VERSION = 1 as const;\n\nexport type ServiceKind =\n\t| \"decoder\"\n\t| \"subgraph-runtime\"\n\t| \"webhook-runtime\"\n\t| \"protocol-producer\";\n\nexport type CanonicalInput = ArchiveDataset;\n\nexport interface ExternalInput {\n\t/** Short name for the dependency: `stacks-node-rpc`, `hiro-extended-api`,\n\t *  `observer-journal`, `contract-source-registry`, etc. */\n\tname: string;\n\t/** Why the archive can't provide this. One sentence. */\n\treason: string;\n\t/** Where an operator can source it. `bundled-stacks`, `own-node`,\n\t *  `hiro-public-api`, `re-derive-from-events`, `manual-import`. */\n\tsource: string;\n\t/** True if the archive supplies enough to re-derive this dependency\n\t *  without a live external source. Rare but happens (e.g. contract ABIs\n\t *  derivable from `transactions.raw_tx`). */\n\trebuildable_from_archive: boolean;\n}\n\nexport interface Output {\n\t/** Table/topic/queue name in the operator's runtime. */\n\ttarget: string;\n\t/** How the operator verifies this output — e.g. `row-count`,\n\t *  `semantic-digest`, `effect-manifest`, `dlq-empty`. */\n\tverification: string;\n}\n\nexport interface ServiceManifest {\n\tname: string;\n\tkind: ServiceKind;\n\tdescription: string;\n\tcanonical_inputs: CanonicalInput[];\n\texternal_inputs: ExternalInput[];\n\toutputs: Output[];\n\t/**\n\t * Explicit judgment on whether R2 alone can rebuild this service.\n\t * `false` REQUIRES `external_inputs` to name what else is needed.\n\t */\n\tr2_alone_can_rebuild: boolean;\n\t/** When true, this service can only replay forward from a starting point\n\t *  the operator supplies (e.g. a webhook consumer's last-delivered id).\n\t *  Publish it, don't hide it. */\n\trequires_operator_state: boolean;\n\t/** Repo path where the service lives, for cross-reference. */\n\tsource_path: string;\n}\n\nexport interface DerivedStageReport {\n\tschema_version: typeof DERIVED_STAGE_REPORT_SCHEMA_VERSION;\n\tgenerated_at: string;\n\tservices: ServiceManifest[];\n}\n\n/**\n * Compose a report from a set of registered manifests. Sorts services by name\n * so the JSON is stable across runs — a re-emit with unchanged manifests\n * produces byte-identical output, so an operator can diff two archive\n * publish cycles cleanly.\n */\nexport function buildDerivedStageReport(\n\tmanifests: ReadonlyArray<ServiceManifest>,\n\tgeneratedAt: string,\n): DerivedStageReport {\n\tconst sorted = [...manifests].sort((a, b) => a.name.localeCompare(b.name));\n\treturn {\n\t\tschema_version: DERIVED_STAGE_REPORT_SCHEMA_VERSION,\n\t\tgenerated_at: generatedAt,\n\t\tservices: sorted,\n\t};\n}\n\nexport type ManifestValidationIssue = {\n\tname: string;\n\tproblem: string;\n};\n\n/**\n * Enforce the invariants the report shape leaves ambiguous:\n *   - `r2_alone_can_rebuild: false` REQUIRES at least one external input.\n *     Otherwise the report claims a rebuild gap it can't explain, which is\n *     worse than declaring nothing.\n *   - Names are unique — a duplicate is almost always a copy-paste that\n *     silently shadows the earlier declaration.\n *   - Every canonical_input is a real dataset name.\n */\nexport function validateServiceManifests(\n\tmanifests: ReadonlyArray<ServiceManifest>,\n): ManifestValidationIssue[] {\n\tconst issues: ManifestValidationIssue[] = [];\n\tconst seen = new Set<string>();\n\tconst canonicalDatasets: ReadonlySet<CanonicalInput> = new Set(\n\t\tARCHIVE_DATASETS,\n\t);\n\tfor (const m of manifests) {\n\t\tif (seen.has(m.name)) {\n\t\t\tissues.push({ name: m.name, problem: \"duplicate service name\" });\n\t\t}\n\t\tseen.add(m.name);\n\t\tif (!m.r2_alone_can_rebuild && m.external_inputs.length === 0) {\n\t\t\tissues.push({\n\t\t\t\tname: m.name,\n\t\t\t\tproblem:\n\t\t\t\t\t\"declares r2_alone_can_rebuild=false but lists no external_inputs — the gap is undocumented\",\n\t\t\t});\n\t\t}\n\t\tfor (const input of m.canonical_inputs) {\n\t\t\tif (!canonicalDatasets.has(input)) {\n\t\t\t\tissues.push({\n\t\t\t\t\tname: m.name,\n\t\t\t\t\tproblem: `canonical_inputs contains unknown dataset: ${String(input)}`,\n\t\t\t\t});\n\t\t\t}\n\t\t}\n\t}\n\treturn issues;\n}\n"
   ],
-  "mappings": ";AA8BO,IAAM,sCAAsC;AAiE5C,SAAS,uBAAuB,CACtC,WACA,aACqB;AAAA,EACrB,MAAM,SAAS,CAAC,GAAG,SAAS,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,KAAK,cAAc,EAAE,IAAI,CAAC;AAAA,EACzE,OAAO;AAAA,IACN,gBAAgB;AAAA,IAChB,cAAc;AAAA,IACd,UAAU;AAAA,EACX;AAAA;AAiBM,SAAS,wBAAwB,CACvC,WAC4B;AAAA,EAC5B,MAAM,SAAoC,CAAC;AAAA,EAC3C,MAAM,OAAO,IAAI;AAAA,EACjB,MAAM,oBAAiD,IAAI,IAAI;AAAA,IAC9D;AAAA,IACA;AAAA,IACA;AAAA,EACD,CAAC;AAAA,EACD,WAAW,KAAK,WAAW;AAAA,IAC1B,IAAI,KAAK,IAAI,EAAE,IAAI,GAAG;AAAA,MACrB,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,SAAS,yBAAyB,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,IAAI,EAAE,IAAI;AAAA,IACf,IAAI,CAAC,EAAE,wBAAwB,EAAE,gBAAgB,WAAW,GAAG;AAAA,MAC9D,OAAO,KAAK;AAAA,QACX,MAAM,EAAE;AAAA,QACR,SACC;AAAA,MACF,CAAC;AAAA,IACF;AAAA,IACA,WAAW,SAAS,EAAE,kBAAkB;AAAA,MACvC,IAAI,CAAC,kBAAkB,IAAI,KAAK,GAAG;AAAA,QAClC,OAAO,KAAK;AAAA,UACX,MAAM,EAAE;AAAA,UACR,SAAS,8CAA8C,OAAO,KAAK;AAAA,QACpE,CAAC;AAAA,MACF;AAAA,IACD;AAAA,EACD;AAAA,EACA,OAAO;AAAA;",
-  "debugId": "664A04684A589A7B64756E2164756E21",
+  "mappings": ";AAYO,IAAM,mBAAmB;AAAA,EAC/B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACD;AAKO,SAAS,gBAAgB,CAAC,OAAwC;AAAA,EACxE,OAAQ,iBAAuC,SAAS,KAAK;AAAA;AAW9D,IAAM,mBAAoD;AAAA,EACzD,QAAQ;AAAA,EACR,cAAc;AAAA,EACd,QAAQ;AAAA,EACR,cAAc;AACf;AAEO,SAAS,iBAAiB,CAAC,SAAkC;AAAA,EACnE,OAAO,iBAAiB;AAAA;AAQlB,SAAS,iBAAiB,CAChC,UAIA,SACU;AAAA,EACV,IAAI,CAAC,kBAAkB,OAAO;AAAA,IAAG,OAAO;AAAA,EACxC,QACE,SAAS,cAAc,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,YAAY,OAAO,MAC5D,SAAS,sBAAsB,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,YAAY,OAAO;AAAA;AAQhE,SAAS,mBAAmB,CAClC,SAC4B;AAAA,EAC5B,QAAQ;AAAA,SACF;AAAA,MACJ,OAAO;AAAA,SACH;AAAA,SACA;AAAA,SACA;AAAA,MACJ,OAAO;AAAA;AAAA,MAEP,OAAO,mBAAmB,OAAO;AAAA;AAAA;AAI7B,SAAS,kBAAkB,CAAC,SAAuB;AAAA,EACzD,MAAM,IAAI,MAAM,8BAA8B,OAAO,OAAO,GAAG;AAAA;;;ACpDzD,IAAM,sCAAsC;AAiE5C,SAAS,uBAAuB,CACtC,WACA,aACqB;AAAA,EACrB,MAAM,SAAS,CAAC,GAAG,SAAS,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,KAAK,cAAc,EAAE,IAAI,CAAC;AAAA,EACzE,OAAO;AAAA,IACN,gBAAgB;AAAA,IAChB,cAAc;AAAA,IACd,UAAU;AAAA,EACX;AAAA;AAiBM,SAAS,wBAAwB,CACvC,WAC4B;AAAA,EAC5B,MAAM,SAAoC,CAAC;AAAA,EAC3C,MAAM,OAAO,IAAI;AAAA,EACjB,MAAM,oBAAiD,IAAI,IAC1D,gBACD;AAAA,EACA,WAAW,KAAK,WAAW;AAAA,IAC1B,IAAI,KAAK,IAAI,EAAE,IAAI,GAAG;AAAA,MACrB,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,SAAS,yBAAyB,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,IAAI,EAAE,IAAI;AAAA,IACf,IAAI,CAAC,EAAE,wBAAwB,EAAE,gBAAgB,WAAW,GAAG;AAAA,MAC9D,OAAO,KAAK;AAAA,QACX,MAAM,EAAE;AAAA,QACR,SACC;AAAA,MACF,CAAC;AAAA,IACF;AAAA,IACA,WAAW,SAAS,EAAE,kBAAkB;AAAA,MACvC,IAAI,CAAC,kBAAkB,IAAI,KAAK,GAAG;AAAA,QAClC,OAAO,KAAK;AAAA,UACX,MAAM,EAAE;AAAA,UACR,SAAS,8CAA8C,OAAO,KAAK;AAAA,QACpE,CAAC;AAAA,MACF;AAAA,IACD;AAAA,EACD;AAAA,EACA,OAAO;AAAA;",
+  "debugId": "4E289EF15D65D38E64756E2164756E21",
   "names": []
 }

package/dist/src/archive/per-height-digest.d.ts

@@ -32,6 +32,7 @@
 *   - blocks:       (height asc)
 *   - transactions: (block_height asc, tx_index asc, tx_id asc)
 *   - events:       (block_height asc, event_index asc, tx_id asc)
+*   - state_writes: (block_height asc, ordinal asc)
 * Per-partition rollup is `sha256(concat(row_digest_bytes))` in stream order.
 *
 * Test vectors under `packages/shared/src/archive/__fixtures__/semantic-digest/`

package/dist/src/archive/per-height-digest.js

@@ -87,15 +87,26 @@ function encodeEvent(row) {
     jsonScalar(row.data)
   ].join(FIELD_SEP);
 }
+function encodeStateWrite(row) {
+  return [
+    scalar(row.block_height),
+    scalar(row.ordinal),
+    scalar(row.tx_index),
+    scalar(row.key),
+    scalar(row.value_hex)
+  ].join(FIELD_SEP);
+}
 var semanticDigest = {
   v1: {
     spec: SEMANTIC_DIGEST_SPEC_V1,
     encodeBlock,
     encodeTransaction,
     encodeEvent,
+    encodeStateWrite,
     block: (row) => sha256Hex(textEncoder.encode(encodeBlock(row))),
     transaction: (row) => sha256Hex(textEncoder.encode(encodeTransaction(row))),
-    event: (row) => sha256Hex(textEncoder.encode(encodeEvent(row)))
+    event: (row) => sha256Hex(textEncoder.encode(encodeEvent(row))),
+    stateWrite: (row) => sha256Hex(textEncoder.encode(encodeStateWrite(row)))
   }
 };
 
@@ -176,5 +187,5 @@ export {
   PerHeightDigestAccumulator
 };
 
-//# debugId=4E237BBF53E5627D64756E2164756E21
+//# debugId=93A25676A491D51564756E2164756E21
 //# sourceMappingURL=per-height-digest.js.map