Publish v0.8.12

Author: platypii

CHANGELOG.md

@@ -1,5 +1,9 @@
 # Icebird Changelog
 
+## [0.8.12]
+ - Add `scanColumn` streaming primitive to the Iceberg SQL data source
+ - Fix commit crash when serializing metadata with snapshot ids above 2^53
+
 ## [0.8.11]
  - Fix pushed-down `where` combined with `limit`/`offset` silently dropping matching rows
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "icebird",
-  "version": "0.8.11",
+  "version": "0.8.12",
   "description": "Apache Iceberg client for javascript",
   "author": "Hyperparam",
   "homepage": "https://hyperparam.app",
@@ -49,16 +49,16 @@
     "test": "vitest run"
   },
   "dependencies": {
-    "hyparquet": "1.26.1",
+    "hyparquet": "1.26.2",
     "hyparquet-compressors": "1.1.1",
     "hyparquet-writer": "0.16.1",
-    "squirreling": "0.12.24"
+    "squirreling": "0.12.25"
   },
   "devDependencies": {
-    "@types/node": "26.0.0",
+    "@types/node": "26.1.0",
     "@vitest/coverage-v8": "4.1.9",
     "eslint": "9.39.4",
-    "eslint-plugin-jsdoc": "63.0.7",
+    "eslint-plugin-jsdoc": "63.0.10",
     "typescript": "6.0.3",
     "vitest": "4.1.9"
   }

test/sql/icebergDataSource.test.js

@@ -692,12 +692,11 @@ describe.concurrent('icebergDataSource scanColumn', () => {
     expect(viaHook).toHaveLength(5)
   })
 
-  it('characterizes: N aggregates on one column re-scan it N times (no coalescing)', async () => {
-    // Current behavior, pinned for visibility — NOT an endorsement. squirreling
-    // runs the streaming-aggregate fast path once per aggregate, so five
-    // aggregates over one column re-scan the column five times. Coalescing them
-    // into one pass is an upstream squirreling opportunity, not an icebird bug;
-    // this characterization makes a future fix a visible diff here.
+  it('characterizes: N aggregates on one column coalesce into a single scan', async () => {
+    // Current behavior, pinned for visibility. squirreling coalesces the
+    // streaming-aggregate fast path, so five aggregates over one column share a
+    // single scan of that column. (Earlier squirreling releases re-scanned once
+    // per aggregate; this characterization tracks the behavior as a visible diff.)
     const source = await icebergDataSource({ tableUrl, resolver, metadataFileName: 'v2.metadata.json' })
     const baseScanColumn = source.scanColumn
     if (!baseScanColumn) throw new Error('scanColumn not implemented')
@@ -718,6 +717,6 @@ describe.concurrent('icebergDataSource scanColumn', () => {
       query: 'SELECT COUNT("Popularity Rank") AS c, MIN("Popularity Rank") AS mn, MAX("Popularity Rank") AS mx, SUM("Popularity Rank") AS s, AVG("Popularity Rank") AS a FROM bunnies',
     }))
 
-    expect(scanColumnCalls).toBe(5)
+    expect(scanColumnCalls).toBe(1)
   })
 })