feat: recompute computed column rows whose inputs changed

refresh_column fills nulls, so once a row has a value nothing revisits it: an update to one of its inputs, or a definition change, leaves the computed value stale for good. This stamps the column's field metadata with the definition it was computed under and a per-fragment signature of the input storage it was read from (input data files and overlays; not the deletion file, since a delete changes no surviving value). A refresh recomputes every live row of a fragment whose stamp disagrees with the manifest, then records what it computed from in a second commit after the fill. A compacted fragment inherits freshness through the Rewrite lineage when every fragment it was built from was signed, or was appended since the stamp, never had an input moved, and left its rows of the product unfilled (a raw append may supply a value; the product's data is the evidence, and the null fill covers those rows); otherwise it recomputes. A column declared before the stamps existed keeps the null-fill contract on its first refresh, which enrolls it as it stood. The stamp is a metadata-only commit on the computed columns, so a materialized view's drift check treats it like the fill. The core lives in `table::freshness` so a remote refresh can share the contract.
Bump version: 0.39.0-beta.5 → 0.39.0-beta.6
2026-09-11 15:52:17 +00:00 · 2026-09-11 04:52:16 +00:00 · 2026-09-08 12:14:12 +00:00 · 2026-09-08 05:12:45 -07:00 · 2026-09-08 12:03:46 +00:00 · 2026-09-08 04:38:22 -07:00
35 changed files with 3445 additions and 249 deletions
@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.39.0-beta.4"
+current_version = "0.39.0-beta.6"
 parse = """(?x)
    (?P<major>0|[1-9]\\d*)\\.
    (?P<minor>0|[1-9]\\d*)\\.
@@ -3526,8 +3526,8 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"

 [[package]]
 name = "fsst"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "rand 0.9.5",
@@ -4886,8 +4886,8 @@ checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"

 [[package]]
 name = "lance"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arc-swap",
 "arrow",
@@ -4959,8 +4959,8 @@ dependencies = [

 [[package]]
 name = "lance-arrow"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4982,7 +4982,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-scalar"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4996,7 +4996,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-stats"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5005,8 +5005,8 @@ dependencies = [

 [[package]]
 name = "lance-bitpacking"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrayref",
 "crunchy",
@@ -5016,8 +5016,8 @@ dependencies = [

 [[package]]
 name = "lance-core"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5054,8 +5054,8 @@ dependencies = [

 [[package]]
 name = "lance-datafusion"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5085,8 +5085,8 @@ dependencies = [

 [[package]]
 name = "lance-datagen"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5103,8 +5103,8 @@ dependencies = [

 [[package]]
 name = "lance-derive"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -5113,8 +5113,8 @@ dependencies = [

 [[package]]
 name = "lance-encoding"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-arith",
 "arrow-array",
@@ -5147,8 +5147,8 @@ dependencies = [

 [[package]]
 name = "lance-file"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-arith",
 "arrow-array",
@@ -5179,8 +5179,8 @@ dependencies = [

 [[package]]
 name = "lance-index"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arc-swap",
 "arrow",
@@ -5244,8 +5244,8 @@ dependencies = [

 [[package]]
 name = "lance-index-core"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5267,8 +5267,8 @@ dependencies = [

 [[package]]
 name = "lance-io"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5308,8 +5308,8 @@ dependencies = [

 [[package]]
 name = "lance-linalg"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5323,8 +5323,8 @@ dependencies = [

 [[package]]
 name = "lance-namespace"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow",
 "async-trait",
@@ -5338,8 +5338,8 @@ dependencies = [

 [[package]]
 name = "lance-namespace-impls"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow",
 "arrow-ipc",
@@ -5392,8 +5392,8 @@ dependencies = [

 [[package]]
 name = "lance-select"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5407,8 +5407,8 @@ dependencies = [

 [[package]]
 name = "lance-table"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5448,8 +5448,8 @@ dependencies = [

 [[package]]
 name = "lance-testing"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5462,8 +5462,8 @@ dependencies = [

 [[package]]
 name = "lance-tokenizer"
-version = "12.0.0-beta.14"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.14#a8a101774a1c9647065cc60137094feadbe55296"
+version = "12.0.0-beta.15"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.15#271a155ba505fd8e1094c095d4ce356707e93866"
 dependencies = [
 "frostem",
 "icu_segmenter",
@@ -5476,7 +5476,7 @@ dependencies = [

 [[package]]
 name = "lancedb"
-version = "0.39.0-beta.4"
+version = "0.39.0-beta.6"
 dependencies = [
 "ahash",
 "anyhow",
@@ -5553,6 +5553,7 @@ dependencies = [
 "serde_json",
 "serde_with",
 "serial_test",
+ "sha2 0.10.9",
 "snafu 0.8.9",
 "tempfile",
 "test-log",
@@ -5567,7 +5568,7 @@ dependencies = [

 [[package]]
 name = "lancedb-nodejs"
-version = "0.39.0-beta.4"
+version = "0.39.0-beta.6"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5592,7 +5593,7 @@ dependencies = [

 [[package]]
 name = "lancedb-python"
-version = "0.39.0-beta.4"
+version = "0.39.0-beta.6"
 dependencies = [
 "arrow",
 "async-trait",
@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"

 [workspace.dependencies]
-lance = { "version" = "=12.0.0-beta.14", default-features = false, "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=12.0.0-beta.14", default-features = false, "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=12.0.0-beta.14", default-features = false, "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=12.0.0-beta.14", "tag" = "v12.0.0-beta.14", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=12.0.0-beta.15", default-features = false, "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=12.0.0-beta.15", default-features = false, "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=12.0.0-beta.15", default-features = false, "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=12.0.0-beta.15", "tag" = "v12.0.0-beta.15", "git" = "https://github.com/lance-format/lance.git" }
 lancedb = { path = "rust/lancedb", default-features = false }
 ahash = "0.8"
 # Note that this one does not include pyarrow
@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
    <groupId>com.lancedb</groupId>
    <artifactId>lancedb-core</artifactId>
-    <version>0.39.0-beta.4</version>
+    <version>0.39.0-beta.6</version>
 </dependency>
 ```

@@ -74,10 +74,10 @@ now: the column is committed with no values, and rows get them from
 [Table#refreshColumn](Table.md#refreshcolumn). Declaring one therefore costs the same on a
 large table as on an empty one.

-A refresh does not revisit rows it has already filled, so mutating an
-input leaves the value computed at fill time; recomputing means dropping
-the column and declaring it again. While a declaration reads a column,
-that column cannot be renamed, retyped or dropped.
+A refresh also recomputes the rows whose inputs changed since they were
+computed, so a mutated input is reflected by the next refresh. While a
+declaration reads a column, that column cannot be renamed, retyped or
+dropped.

 On LanceDB Cloud and Enterprise the expression is planned by the
 server, and the refresh runs as a server job -- see
@@ -854,10 +854,10 @@ abstract refreshColumn(column): Promise<RefreshColumnResult>

 Fill the rows of a computed column that hold no value yet.

-Rows appended since the last refresh are filled by the next one; rows
-already filled are left as they are, so the call is idempotent and does
-not observe a mutated input. Local tables only: a remote refresh runs
-as a server job, through [Table#refreshColumnAsync](Table.md#refreshcolumnasync).
+Rows appended since the last refresh are filled by the next one, and
+rows whose inputs changed since they were computed are recomputed;
+everything else is left as it is. Local tables only: a remote refresh
+runs as a server job, through [Table#refreshColumnAsync](Table.md#refreshcolumnasync).

 #### Parameters

@@ -8,7 +8,7 @@
    <parent>
      <groupId>com.lancedb</groupId>
      <artifactId>lancedb-parent</artifactId>
-      <version>0.39.0-beta.4</version>
+      <version>0.39.0-beta.6</version>
      <relativePath>../pom.xml</relativePath>
    </parent>

@@ -6,7 +6,7 @@

    <groupId>com.lancedb</groupId>
    <artifactId>lancedb-parent</artifactId>
-    <version>0.39.0-beta.4</version>
+    <version>0.39.0-beta.6</version>
    <packaging>pom</packaging>
    <name>${project.artifactId}</name>
    <description>LanceDB Java SDK Parent POM</description>
@@ -28,7 +28,7 @@
    <properties>
        <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
        <arrow.version>15.0.0</arrow.version>
-        <lance-core.version>12.0.0-beta.14</lance-core.version>
+        <lance-core.version>12.0.0-beta.15</lance-core.version>
        <spotless.skip>false</spotless.skip>
        <spotless.version>2.30.0</spotless.version>
        <spotless.java.googlejavaformat.version>1.7</spotless.java.googlejavaformat.version>
@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.39.0-beta.4"
+version = "0.39.0-beta.6"
 publish = false
 license.workspace = true
 description.workspace = true
@@ -281,7 +281,7 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
        numIndices: 0,
        numRows: 3,
        // Full on-disk size of the two data files, footers and metadata included.
-        totalBytes: 684,
+        totalBytes: 550,
      });

      // Index files count toward totalBytes too (only deletion files and
@@ -289,7 +289,7 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
      await table.createIndex("id", { config: Index.btree() });
      const statsWithIndex = await table.stats();
      expect(statsWithIndex.numIndices).toBe(1);
-      expect(statsWithIndex.totalBytes).toBeGreaterThan(684);
+      expect(statsWithIndex.totalBytes).toBeGreaterThan(550);
    });

    it("should overwrite data if asked", async () => {
@@ -542,10 +542,10 @@ export abstract class Table {
   * {@link Table#refreshColumn}. Declaring one therefore costs the same on a
   * large table as on an empty one.
   *
-   * A refresh does not revisit rows it has already filled, so mutating an
-   * input leaves the value computed at fill time; recomputing means dropping
-   * the column and declaring it again. While a declaration reads a column,
-   * that column cannot be renamed, retyped or dropped.
+   * A refresh also recomputes the rows whose inputs changed since they were
+   * computed, so a mutated input is reflected by the next refresh. While a
+   * declaration reads a column, that column cannot be renamed, retyped or
+   * dropped.
   *
   * On LanceDB Cloud and Enterprise the expression is planned by the
   * server, and the refresh runs as a server job -- see
@@ -576,10 +576,10 @@ export abstract class Table {
  /**
   * Fill the rows of a computed column that hold no value yet.
   *
-   * Rows appended since the last refresh are filled by the next one; rows
-   * already filled are left as they are, so the call is idempotent and does
-   * not observe a mutated input. Local tables only: a remote refresh runs
-   * as a server job, through {@link Table#refreshColumnAsync}.
+   * Rows appended since the last refresh are filled by the next one, and
+   * rows whose inputs changed since they were computed are recomputed;
+   * everything else is left as it is. Local tables only: a remote refresh
+   * runs as a server job, through {@link Table#refreshColumnAsync}.
   * @param {string} column The name of the computed column to fill.
   * @returns {Promise<RefreshColumnResult>} A promise that resolves to the
   * number of rows filled and the new version number of the table.
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.39.0-beta.4",
+	"version": "0.39.0-beta.6",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.39.0-beta.4",
+	"version": "0.39.0-beta.6",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.39.0-beta.4",
+	"version": "0.39.0-beta.6",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.39.0-beta.4",
+	"version": "0.39.0-beta.6",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.39.0-beta.4",
+	"version": "0.39.0-beta.6",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",
@@ -1,6 +1,6 @@
 {
  "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.39.0-beta.4",
+  "version": "0.39.0-beta.6",
  "os": [
    "win32"
  ],
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.39.0-beta.4",
+	"version": "0.39.0-beta.6",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",
@@ -11,7 +11,7 @@
    "ann"
  ],
  "private": false,
-  "version": "0.39.0-beta.4",
+  "version": "0.39.0-beta.6",
  "main": "dist/index.js",
  "exports": {
    ".": "./dist/index.js",
@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.39.0-beta.4"
+version = "0.39.0-beta.6"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"
@@ -2188,10 +2188,10 @@ class Table(ABC):
            Declaring one therefore costs the same on a large table as on an
            empty one.

-            A refresh does not revisit rows it has already filled, so mutating
-            an input leaves the value computed at fill time; recomputing means
-            dropping the column and declaring it again. While a declaration
-            reads a column, that column cannot be renamed, retyped or dropped.
+            A refresh also recomputes the rows whose inputs changed since they
+            were computed, so a mutated input is reflected by the next refresh.
+            While a declaration reads a column, that column cannot be renamed,
+            retyped or dropped.

            On LanceDB Cloud and Enterprise the expression is planned by the
            server, and the refresh runs as a server job -- see
@@ -2211,7 +2211,7 @@ class Table(ABC):
        >>> table.add_columns(computed={"doubled": "x * 2"})
        AddColumnsResult(version=2)
        >>> table.refresh_column("doubled")
-        RefreshColumnResult(rows_filled=2, version=3)
+        RefreshColumnResult(rows_filled=2, version=4)
        >>> table.to_arrow().sort_by("x").to_pandas()
           x  doubled
        0  1        2
@@ -2225,8 +2225,8 @@ class Table(ABC):

        Declared with ``add_columns(computed=...)``, a column starts empty and
        gets its values here. Rows appended since the last refresh are filled
-        by the next one; rows already filled are left as they are, so the call
-        is idempotent and does not observe a mutated input.
+        by the next one, and rows whose inputs changed since they were computed
+        are recomputed; everything else is left as it is.

        Local tables only: a remote refresh runs as a server job, through
        [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
@@ -4318,13 +4318,14 @@ class LanceTable(Table):
        return LOOP.run(self._table.add_columns(transforms, computed=computed))

    def refresh_column(self, column: str) -> "RefreshColumnResult":
-        """Fill a computed column's unfilled rows. See
+        """Fill a computed column's unfilled rows and recompute those whose
+        inputs changed. See
        [`AsyncTable.refresh_column`][lancedb.AsyncTable.refresh_column]."""
        return LOOP.run(self._table.refresh_column(column))

    def refresh_column_async(self, column: str) -> Job[RefreshColumnJobResult]:
-        """Fill a computed column's unfilled rows, returning a handle to the
-        refresh job. See
+        """Fill a computed column's unfilled rows and recompute those whose
+        inputs changed, returning a handle to the refresh job. See
        [`Table.refresh_column_async`][lancedb.table.Table.refresh_column_async].
        """
        return Job(LOOP.run(self._table.refresh_column_async(column)))
@@ -6312,10 +6313,10 @@ class AsyncTable:
            them from
            [`refresh_column`][lancedb.table.AsyncTable.refresh_column].

-            A refresh does not revisit rows it has already filled, so mutating
-            an input leaves the value computed at fill time. While a
-            declaration reads a column, that column cannot be renamed, retyped
-            or dropped.
+            A refresh also recomputes the rows whose inputs changed since they
+            were computed, so a mutated input is reflected by the next refresh.
+            While a declaration reads a column, that column cannot be renamed,
+            retyped or dropped.

            On LanceDB Cloud and Enterprise the expression is planned by
            the server. Cannot be combined with ``transforms``.
@@ -6377,8 +6378,8 @@ class AsyncTable:

        Declared with ``add_columns(computed=...)``, a column starts empty and
        gets its values here. Rows appended since the last refresh are filled
-        by the next one; rows already filled are left as they are, so the call
-        is idempotent and does not observe a mutated input.
+        by the next one, and rows whose inputs changed since they were computed
+        are recomputed; everything else is left as it is.

        Local tables only: a remote refresh runs as a server job, through
        [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
@@ -297,7 +297,10 @@ def test_blob_v2_projection_sources_use_typed_column_name():


 def _legacy_v1_table(name):
-    db = lancedb.connect("memory:///")
+    # Legacy v1 blob columns are only writable at file version <= 2.1.
+    db = lancedb.connect(
+        "memory:///", storage_options={"new_table_data_storage_version": "2.1"}
+    )
    schema = pa.schema(
        [
            pa.field("id", pa.int64()),
@@ -193,7 +193,13 @@ class TestNamespaceConnection:
            ),
        )

-        table = db.create_table("blob_table", data, namespace_path=["test_ns"])
+        # Legacy v1 blob columns are only writable at file version <= 2.1.
+        table = db.create_table(
+            "blob_table",
+            data,
+            namespace_path=["test_ns"],
+            storage_options={"new_table_data_storage_version": "2.1"},
+        )
        df = table.to_pandas(blob_mode="lazy").sort_values("id")

        blob = df["blob"].iloc[0]
@@ -40,6 +40,10 @@ from utils import exception_output
 from importlib.util import find_spec


+# Legacy v1 blob columns are only writable at file version <= 2.1.
+LEGACY_BLOB_STORAGE_OPTIONS = {"new_table_data_storage_version": "2.1"}
+
+
 def _blob_query_data():
    return pa.table(
        {
@@ -119,13 +123,17 @@ def _assert_blob_bytes_projection(df):

 def _blob_query_table(db, name, blob_schema):
    if blob_schema == "v1":
-        return db.create_table(name, _blob_query_data())
+        return db.create_table(
+            name, _blob_query_data(), storage_options=LEGACY_BLOB_STORAGE_OPTIONS
+        )
    return _create_blob_v2_query_table(db, name)


 async def _blob_query_table_async(db, name, blob_schema):
    if blob_schema == "v1":
-        return await db.create_table(name, _blob_query_data())
+        return await db.create_table(
+            name, _blob_query_data(), storage_options=LEGACY_BLOB_STORAGE_OPTIONS
+        )
    return await _create_blob_v2_query_table_async(db, name)


@@ -275,7 +283,9 @@ async def test_query_to_pandas_kwargs(table, table_async):
 def test_plain_scan_query_to_pandas_blob_modes(tmp_db, blob_mode):
    pytest.importorskip("lance")
    table = tmp_db.create_table(
-        f"test_query_to_pandas_blob_{blob_mode}", _blob_query_data()
+        f"test_query_to_pandas_blob_{blob_mode}",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
    )

    df = (
@@ -322,7 +332,9 @@ def test_plain_scan_query_to_pandas_blob_mode_does_not_collect_arrow(
 ):
    pytest.importorskip("lance")
    table = tmp_db.create_table(
-        "test_query_to_pandas_blob_no_arrow_collect", _blob_query_data()
+        "test_query_to_pandas_blob_no_arrow_collect",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
    )
    query = table.search().where("id = 1").select(["id", "blob"])

@@ -347,7 +359,9 @@ def test_plain_scan_query_to_pandas_blob_descriptions_flatten_uses_scanner(
 ):
    pytest.importorskip("lance")
    table = tmp_db.create_table(
-        "test_query_to_pandas_blob_desc_flatten", _blob_query_data()
+        "test_query_to_pandas_blob_desc_flatten",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
    )
    query = table.search().where("id = 1").select(["id", "blob"])

@@ -365,7 +379,11 @@ def test_plain_scan_query_to_pandas_blob_descriptions_flatten_uses_scanner(
 def test_plain_scan_query_to_pandas_scanner_state(tmp_db):
    pytest.importorskip("lance")
    data = _blob_query_data()
-    table = tmp_db.create_table("test_query_to_pandas_scanner_state", data.slice(0, 2))
+    table = tmp_db.create_table(
+        "test_query_to_pandas_scanner_state",
+        data.slice(0, 2),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
+    )
    table.add(data.slice(2, 2))

    fragments = table.to_lance().get_fragments()
@@ -400,7 +418,9 @@ def test_plain_scan_query_to_pandas_scanner_state(tmp_db):
 async def test_async_plain_scan_query_to_pandas_blob_projection(tmp_db_async):
    pytest.importorskip("lance")
    table = await tmp_db_async.create_table(
-        "test_async_query_to_pandas_blob_projection", _blob_query_data()
+        "test_async_query_to_pandas_blob_projection",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
    )

    lazy_df = await (
@@ -452,7 +472,9 @@ async def test_async_plain_scan_query_to_pandas_blob_mode_does_not_collect_arrow
 ):
    pytest.importorskip("lance")
    table = await tmp_db_async.create_table(
-        "test_async_query_to_pandas_blob_no_arrow_collect", _blob_query_data()
+        "test_async_query_to_pandas_blob_no_arrow_collect",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
    )
    query = table.query().where("id = 1").select(["id", "blob"])

@@ -474,7 +496,11 @@ async def test_async_plain_scan_query_to_pandas_blob_mode_does_not_collect_arrow

 def test_vector_query_to_pandas_blob_mode_requires_native_path(tmp_db):
    pytest.importorskip("lance")
-    table = tmp_db.create_table("test_vector_query_blob_mode", _blob_query_data())
+    table = tmp_db.create_table(
+        "test_vector_query_blob_mode",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
+    )

    with pytest.raises(RuntimeError, match="Lance native pandas conversion"):
        table.search([1.0, 0.0]).select(["blob", "vector"]).limit(1).to_pandas(
@@ -485,7 +511,9 @@ def test_vector_query_to_pandas_blob_mode_requires_native_path(tmp_db):
 def test_vector_query_to_pandas_blob_descriptions_requires_plain_scan(tmp_db):
    pytest.importorskip("lance")
    table = tmp_db.create_table(
-        "test_vector_query_blob_descriptions", _blob_query_data()
+        "test_vector_query_blob_descriptions",
+        _blob_query_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
    )

    with pytest.raises(RuntimeError, match="plain scan query"):
@@ -64,15 +64,23 @@ async def _blob_v2_table_async(db: AsyncConnection, name: str):
    return table


+# Legacy v1 blob columns are only writable at file version <= 2.1.
+LEGACY_BLOB_STORAGE_OPTIONS = {"new_table_data_storage_version": "2.1"}
+
+
 def _blob_table(db: DBConnection, name: str, blob_schema: str):
    if blob_schema == "v1":
-        return db.create_table(name, data=_blob_test_data())
+        return db.create_table(
+            name, data=_blob_test_data(), storage_options=LEGACY_BLOB_STORAGE_OPTIONS
+        )
    return _blob_v2_table(db, name)


 async def _blob_table_async(db: AsyncConnection, name: str, blob_schema: str):
    if blob_schema == "v1":
-        return await db.create_table(name, data=_blob_test_data())
+        return await db.create_table(
+            name, data=_blob_test_data(), storage_options=LEGACY_BLOB_STORAGE_OPTIONS
+        )
    return await _blob_v2_table_async(db, name)


@@ -147,7 +155,11 @@ def test_table_to_pandas_invalid_blob_mode_non_blob_table(tmp_db: DBConnection):
@pytest.mark.parametrize("blob_mode", ["lazy", "bytes", "descriptions"])
 def test_table_to_pandas_blob_modes(tmp_db: DBConnection, blob_mode):
    pytest.importorskip("lance")
-    table = tmp_db.create_table(f"test_to_pandas_blob_{blob_mode}", _blob_test_data())
+    table = tmp_db.create_table(
+        f"test_to_pandas_blob_{blob_mode}",
+        _blob_test_data(),
+        storage_options=LEGACY_BLOB_STORAGE_OPTIONS,
+    )

    df = table.to_pandas(blob_mode=blob_mode)

@@ -3959,7 +3971,7 @@ def test_stats(mem_db: DBConnection):
    print(f"{stats=}")
    assert stats == {
        # Full on-disk size of the data file, footer and metadata included.
-        "total_bytes": 633,
+        "total_bytes": 637,
        "num_rows": 2,
        "num_indices": 0,
        "fragment_stats": {
@@ -4171,13 +4183,14 @@ def test_refresh_column_async_returns_job(tmp_path):
    assert result.rows_failed == 0
    assert result.rows_remaining == 0
    assert result.source_version == 2
-    assert result.published_version == 3
+    # The fill lands at 3; the stamp recording its inputs is published at 4.
+    assert result.published_version == 4
    assert job.status() == "finished"
    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4]

    no_op = table.refresh_column_async("doubled").wait()
    assert no_op.rows_assigned == 0
-    assert no_op.source_version == 3
+    assert no_op.source_version == 4
    assert no_op.published_version is None

    # Bad input raises at the call, not through the job.
@@ -4196,6 +4209,6 @@ async def test_refresh_column_async_job_async_table(tmp_path):
    assert isinstance(result, lancedb.RefreshColumnResult)
    assert result.rows_assigned == 1
    assert result.source_version == 2
-    assert result.published_version == 3
+    assert result.published_version == 4
    assert await job.status() == "finished"
    assert (await table.to_arrow())["tripled"].to_pylist() == [9]
@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.39.0-beta.4"
+version = "0.39.0-beta.6"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -95,13 +95,14 @@ candle-transformers = { version = "0.9.1", optional = true }
 candle-nn = { version = "0.9.1", optional = true }
 tokenizers = { version = "0.19.1", optional = true }
 semver = { workspace = true }
+roaring = "0.11.4"
+sha2 = "0.10"

 [dev-dependencies]
 anyhow = "1"
 lance-testing = { workspace = true }
 tempfile = { workspace = true }
 random_word = { version = "0.4.3", features = ["en"] }
-roaring = "0.11.4"
 tokio = { workspace = true, features = ["io-util", "macros", "net", "test-util"] }
 uuid = { workspace = true }
 walkdir = "2"
@@ -532,10 +532,11 @@ mod tests {
    fn storage_version_bumps_to_v2_2() {
        let mut params = WriteParams::default();
        ensure_blob_storage_version(&blob_schema(), &mut params);
-        assert_eq!(
-            params.data_storage_version.unwrap().resolve(),
-            ConcreteFileVersion::V2_2
-        );
+        let resolved = params
+            .data_storage_version
+            .unwrap_or(LanceFileVersion::Stable)
+            .resolve();
+        assert_eq!(resolved, ConcreteFileVersion::V2_2);
        assert!(!params.enable_stable_row_ids);
    }

@@ -547,10 +548,11 @@ mod tests {
        };
        ensure_blob_storage_version(&blob_schema(), &mut params);
        assert!(params.enable_stable_row_ids);
-        assert_eq!(
-            params.data_storage_version.unwrap().resolve(),
-            ConcreteFileVersion::V2_2
-        );
+        let resolved = params
+            .data_storage_version
+            .unwrap_or(LanceFileVersion::Stable)
+            .resolve();
+        assert_eq!(resolved, ConcreteFileVersion::V2_2);
    }

    #[test]
@@ -3,6 +3,7 @@

 //! Namespace-based database implementation that delegates table management to lance-namespace

+use lance_datafusion::utils::StreamingWriteSource;
 use std::collections::{HashMap, HashSet};
 use std::sync::{Arc, Mutex};

@@ -304,6 +305,10 @@ impl Database for LanceNamespaceDatabase {
    }

    async fn create_table(&self, request: DbCreateTableRequest) -> Result<Arc<dyn BaseTable>> {
+        // Refuse a bad declaration before the namespace records a table.
+        crate::table::computed_columns::ensure_declarations_are_planned(
+            &request.data.arrow_schema(),
+        )?;
        let mut table_id = request.namespace_path.clone();
        table_id.push(request.name.clone());
        let mut existing_table = None;
@@ -24,8 +24,8 @@ use std::time::{SystemTime, UNIX_EPOCH};

 use arrow_array::cast::AsArray;
 use arrow_array::types::UInt64Type;
-use arrow_array::{RecordBatch, UInt64Array};
-use arrow_schema::{Schema as ArrowSchema, SchemaRef};
+use arrow_array::{RecordBatch, UInt64Array, new_null_array};
+use arrow_schema::{FieldRef, Schema as ArrowSchema, SchemaRef};
 use datafusion::common::ScalarValue;
 use datafusion::error::DataFusionError;
 use datafusion::physical_plan::SendableRecordBatchStream;
@@ -34,7 +34,7 @@ use datafusion::prelude::{col, lit};
 use futures::{StreamExt, TryStreamExt};
 use lance::Dataset;
 use lance::dataset::mem_wal::DatasetMemWalExt;
-use lance::dataset::transaction::{Operation, Transaction};
+use lance::dataset::transaction::{Operation, Transaction, UpdateMode};
 use lance::dataset::write::delete::DeleteBuilder;
 use lance::dataset::write::merge_insert::inserted_rows::{
    KeyExistenceFilter, KeyExistenceFilterBuilder, KeyValue,
@@ -51,6 +51,9 @@ use super::{
    definition_to_metadata,
 };
 use crate::database::OpenTableRequest;
+use crate::table::computed_columns::{
+    computed_column_from_field, computed_columns, ensure_declarations_are_planned,
+};
 use crate::table::{NativeTable, NativeTableExt, Table};
 use crate::{Error, Result};

@@ -167,30 +170,52 @@ pub(crate) async fn execute_refresh(
        .map(|p| (p.output.clone(), p.expression.clone()))
        .collect();
    validate_inputs(&source_ds, definition)?;
-    let (replanned, mut planned_fields, _renames) = super::plan(
+    let (replanned, planned_fields, _renames) = super::plan(
        source_schema,
        &definition.source_table,
        &definition.source_namespace,
-        &projections,
+        Some(&projections),
        definition.filter.as_deref(),
        definition.limit,
    )?;
+    let mut planned_fields = planned_fields;
    planned_fields.push(arrow_schema::Field::new(
        SOURCE_ROW_ID_COLUMN,
        arrow_schema::DataType::UInt64,
        false,
    ));
+    // A computed column is not planned from the source: refresh writes it
+    // NULL and its declaration's owner fills it. Its declaration must still
+    // be complete, and it must be able to hold NULL.
    let physical = ArrowSchema::from(view_ds.schema());
-    let planned_shape: Vec<_> = planned_fields
-        .iter()
-        .map(|f| (f.name().clone(), f.data_type().clone(), f.is_nullable()))
-        .collect();
-    let physical_shape: Vec<_> = physical
+    let mut computed = computed_columns(&physical).into_iter().map(|c| c.name);
+    if let Some(name) = computed.by_ref().find(|name| {
+        physical
+            .field_with_name(name)
+            .is_ok_and(|f| !f.is_nullable())
+    }) {
+        return Err(Error::Schema {
+            message: format!(
+                "computed column '{name}' of view '{}' cannot hold NULL; recreate the view",
+                view.name()
+            ),
+        });
+    }
+    ensure_declarations_are_planned(&physical)?;
+    let physical_planned: Vec<&FieldRef> = physical
        .fields()
        .iter()
-        .map(|f| (f.name().clone(), f.data_type().clone(), f.is_nullable()))
+        .filter(|f| computed_column_from_field(f).is_none())
        .collect();
-    if planned_shape != physical_shape {
+    // A projected column that became nullable at the source still fits the
+    // view's nullable field; the reverse would not.
+    let matches = planned_fields.len() == physical_planned.len()
+        && planned_fields.iter().zip(&physical_planned).all(|(e, p)| {
+            e.name() == p.name()
+                && e.data_type() == p.data_type()
+                && (p.is_nullable() || !e.is_nullable())
+        });
+    if !matches {
        return Err(Error::Schema {
            message: format!(
                "the stored definition of view '{}' does not produce this \
@@ -229,11 +254,18 @@ pub(crate) async fn execute_refresh(
        .get(SOURCE_VERSION_TS_META_KEY)
        .and_then(|raw| raw.parse().ok());
    // The watermark speaks only for the view state its refresh left behind;
-    // any other commit on the view since then is drift.
-    let view_intact = metadata
+    // any other commit on the view since then is drift, except a fill of its
+    // computed columns, which rewrites nothing refresh certifies.
+    let recorded_view_version = metadata
        .get(VIEW_VERSION_META_KEY)
-        .and_then(|raw| raw.parse::<u64>().ok())
-        == Some(view_ds.version().version);
+        .and_then(|raw| raw.parse::<u64>().ok());
+    let view_intact = match recorded_view_version {
+        Some(recorded) if recorded == view_ds.version().version => true,
+        Some(recorded) if recorded < view_ds.version().version => {
+            only_computed_rewrites_since(&view_ds, recorded).await?
+        }
+        _ => false,
+    };

    if !full && watermark == Some(source_version) && view_intact && recorded_ts == Some(source_ts) {
        return Ok(RefreshMaterializedViewResult {
@@ -1090,6 +1122,83 @@ struct RowScope {
    limit: Option<u64>,
 }

+/// Whether every commit on the view after `recorded` is a fill of its
+/// computed columns: a column rewrite or data replacement touching only
+/// those fields and neither adding nor removing rows, or the freshness
+/// stamp a fill leaves on them. A version whose transaction cannot be read
+/// is not proven, so it counts as drift.
+async fn only_computed_rewrites_since(view_ds: &Dataset, recorded: u64) -> Result<bool> {
+    // A fill may write any field under a computed column, so the whole
+    // subtree counts, not only the root.
+    let physical = ArrowSchema::from(view_ds.schema());
+    fn subtree(field: &lance_core::datatypes::Field, ids: &mut Vec<u32>) {
+        ids.push(field.id as u32);
+        for child in &field.children {
+            subtree(child, ids);
+        }
+    }
+    let mut computed_fields = Vec::new();
+    for column in computed_columns(&physical) {
+        if let Some(field) = view_ds.schema().field(&column.name) {
+            subtree(field, &mut computed_fields);
+        }
+    }
+    if computed_fields.is_empty() {
+        return Ok(false);
+    }
+    for version in recorded + 1..=view_ds.version().version {
+        let Some(transaction) = view_ds.read_transaction_by_version(version).await? else {
+            return Ok(false);
+        };
+        let fill = match &transaction.operation {
+            Operation::Update {
+                removed_fragment_ids,
+                new_fragments,
+                fields_modified,
+                update_mode: Some(UpdateMode::RewriteColumns),
+                ..
+            } => {
+                removed_fragment_ids.is_empty()
+                    && new_fragments.is_empty()
+                    && !fields_modified.is_empty()
+                    && fields_modified
+                        .iter()
+                        .all(|field| computed_fields.contains(field))
+            }
+            // What `refresh_column` commits for a SQL declaration.
+            Operation::DataReplacement { replacements } => {
+                !replacements.is_empty()
+                    && replacements.iter().all(|group| {
+                        !group.1.fields.is_empty()
+                            && group
+                                .1
+                                .fields
+                                .iter()
+                                .all(|field| computed_fields.contains(&(*field as u32)))
+                    })
+            }
+            // The stamp `refresh_column` writes after its fill (see
+            // `table::freshness`): field metadata on computed columns, no data.
+            Operation::UpdateConfig {
+                config_updates: None,
+                table_metadata_updates: None,
+                schema_metadata_updates: None,
+                field_metadata_updates,
+            } => {
+                !field_metadata_updates.is_empty()
+                    && field_metadata_updates
+                        .keys()
+                        .all(|field| computed_fields.contains(&(*field as u32)))
+            }
+            _ => false,
+        };
+        if !fill {
+            return Ok(false);
+        }
+    }
+    Ok(true)
+}
+
 async fn compute_stream(
    source: &Dataset,
    definition: &MaterializedViewDefinition,
@@ -1158,6 +1267,10 @@ async fn compute_stream(
        let batch = batch.map_err(|e| DataFusionError::External(Box::new(e)))?;
        let mut columns = Vec::with_capacity(out_schema.fields().len());
        for field in out_schema.fields() {
+            if computed_column_from_field(field).is_some() {
+                columns.push(new_null_array(field.data_type(), batch.num_rows()));
+                continue;
+            }
            let name = if field.name() == SOURCE_ROW_ID_COLUMN {
                ROW_ID
            } else {
@@ -2768,7 +2881,7 @@ mod tests {
        let (conn, source) = db_with_source(vec![1]).await;
        let prepared = crate::materialized_view::prepare_declaration(
            &source,
-            &[("x".into(), "x".into()), ("twice".into(), "x * 2".into())],
+            Some(&[("x".into(), "x".into()), ("twice".into(), "x * 2".into())]),
            None,
            None,
        )
@@ -3132,4 +3245,464 @@ mod tests {
        let err = view.refresh().execute().await.unwrap_err();
        assert!(err.to_string().contains("source table 'src'"), "{err}");
    }
+
+    /// A view with a computed column, declared over `people` and refreshed.
+    async fn refreshed_computed_view(conn: &Connection) -> MaterializedView {
+        use crate::materialized_view::tests::{computed_field, people, test_binding};
+        let source = people(conn).await;
+        let view = crate::materialized_view::prepare_declaration(
+            &source,
+            Some(&[
+                ("id".to_string(), "id".to_string()),
+                ("name".to_string(), "name".to_string()),
+            ]),
+            None,
+            None,
+        )
+        .await
+        .unwrap()
+        .with_computed_columns(
+            vec![(2, computed_field("emb", "fb_1", "name"))],
+            &[test_binding("fb_1", "name", "emb")],
+        )
+        .unwrap()
+        .create("v")
+        .await
+        .unwrap();
+        let result = view.refresh().execute().await.unwrap();
+        assert_eq!(result.mode, RefreshMode::Rebuild);
+        view
+    }
+
+    async fn unfilled(view: &MaterializedView) -> usize {
+        view.table()
+            .count_rows(Some("emb IS NULL".to_string()))
+            .await
+            .unwrap()
+    }
+
+    async fn append_people(conn: &Connection, ids: Vec<i32>, names: Vec<&str>) {
+        let batch = record_batch!(("id", Int32, ids), ("name", Utf8, names)).unwrap();
+        conn.open_table("people")
+            .execute()
+            .await
+            .unwrap()
+            .add(batch)
+            .execute()
+            .await
+            .unwrap();
+    }
+
+    /// Commit the fill job's shape on the view: a column rewrite of
+    /// `fields`, touching no rows. The data is left as it is; what matters
+    /// here is how the next refresh classifies the commit.
+    async fn commit_column_rewrite(view: &MaterializedView, fields: &[&str]) {
+        let native = view.table().as_native().unwrap();
+        native.dataset.reload().await.unwrap();
+        let dataset = native.dataset.get().await.unwrap().as_ref().clone();
+        let fields_modified = fields
+            .iter()
+            .map(|name| dataset.schema().field(name).unwrap().id as u32)
+            .collect();
+        let updated_fragments = dataset
+            .get_fragments()
+            .iter()
+            .map(|fragment| fragment.metadata().clone())
+            .collect();
+        let operation = Operation::Update {
+            removed_fragment_ids: Vec::new(),
+            updated_fragments,
+            new_fragments: Vec::new(),
+            fields_modified,
+            compacted_sstables: Vec::new(),
+            fields_for_preserving_frag_bitmap: Vec::new(),
+            update_mode: Some(UpdateMode::RewriteColumns),
+            inserted_rows_filter: None,
+            updated_fragment_offsets: None,
+        };
+        let read_version = dataset.version().version;
+        CommitBuilder::new(WriteDestination::Dataset(Arc::new(dataset)))
+            .execute(Transaction::new(read_version, operation, None))
+            .await
+            .unwrap();
+    }
+
+    /// Refresh never computes a computed column: every row it writes, on a
+    /// rebuild, an append and a rewrite, carries NULL there, and the
+    /// declaration survives all three.
+    #[tokio::test]
+    async fn test_computed_columns_are_written_null_and_kept() {
+        let conn = connect("memory://").execute().await.unwrap();
+        let view = refreshed_computed_view(&conn).await;
+        assert_eq!(unfilled(&view).await, 3);
+
+        append_people(&conn, vec![4, 5], vec!["d", "e"]).await;
+        let result = view.refresh().execute().await.unwrap();
+        assert_eq!(result.mode, RefreshMode::Incremental);
+        assert_eq!(unfilled(&view).await, 5);
+
+        conn.open_table("people")
+            .execute()
+            .await
+            .unwrap()
+            .update()
+            .column("name", "'z'")
+            .only_if("id = 1")
+            .execute()
+            .await
+            .unwrap();
+        view.refresh().execute().await.unwrap();
+        assert_eq!(unfilled(&view).await, 5);
+        assert_eq!(read(view.table(), "id").await, vec![1, 2, 3, 4, 5]);
+
+        let schema = view.table().schema().await.unwrap();
+        assert!(
+            crate::table::computed_columns::function_bindings(&schema)
+                .unwrap()
+                .iter()
+                .any(|b| b.binding_id() == "fb_1"),
+            "the binding envelope was lost"
+        );
+        assert!(
+            computed_column_from_field(schema.field_with_name("emb").unwrap()).is_some(),
+            "the declaration was lost"
+        );
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::NoOp
+        );
+    }
+
+    /// Field metadata on `field` only, the commit shape of the freshness
+    /// stamp `refresh_column` leaves after its fill.
+    async fn commit_field_metadata(view: &MaterializedView, field: &str, key: &str) {
+        let native = view.table().as_native().unwrap();
+        native.dataset.reload().await.unwrap();
+        let mut dataset = native.dataset.get().await.unwrap().as_ref().clone();
+        dataset
+            .update_field_metadata()
+            .update(field, [(key.to_string(), "{}".to_string())])
+            .unwrap()
+            .await
+            .unwrap();
+    }
+
+    /// The stamp is metadata on the computed column and rewrites nothing
+    /// refresh certifies, so it is not drift; the same commit shape on a
+    /// projected column is, like any other write to it.
+    #[tokio::test]
+    async fn test_a_freshness_stamp_is_not_drift() {
+        let conn = connect("memory://").execute().await.unwrap();
+        let view = refreshed_computed_view(&conn).await;
+
+        commit_field_metadata(
+            &view,
+            "emb",
+            crate::table::computed_columns::SOURCE_SIGNATURE_META_KEY,
+        )
+        .await;
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::NoOp
+        );
+
+        commit_field_metadata(&view, "id", "probe").await;
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::Rebuild
+        );
+    }
+
+    /// The fill job's commit rewrites only computed columns. It is the one
+    /// commit on a view that is not drift: the next refresh carries on from
+    /// its watermark instead of rebuilding, which would null what the fill
+    /// just wrote.
+    #[tokio::test]
+    async fn test_a_computed_column_fill_is_not_drift() {
+        let conn = connect("memory://").execute().await.unwrap();
+        let view = refreshed_computed_view(&conn).await;
+
+        commit_column_rewrite(&view, &["emb"]).await;
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::NoOp
+        );
+
+        commit_column_rewrite(&view, &["emb"]).await;
+        append_people(&conn, vec![4], vec!["d"]).await;
+        let result = view.refresh().execute().await.unwrap();
+        assert_eq!(result.mode, RefreshMode::Incremental);
+        assert_eq!(result.rows_written, 1);
+        assert_eq!(read(view.table(), "id").await, vec![1, 2, 3, 4]);
+    }
+
+    /// A column rewrite that reaches a projected column is drift like any
+    /// other write: refresh certifies those columns and must recompute them.
+    #[tokio::test]
+    async fn test_a_rewrite_of_a_projected_column_is_drift() {
+        let conn = connect("memory://").execute().await.unwrap();
+        let view = refreshed_computed_view(&conn).await;
+
+        commit_column_rewrite(&view, &["emb", "name"]).await;
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::Rebuild
+        );
+    }
+
+    /// The declaration contract is checked before any refresh mutation: a
+    /// missing binding envelope and a column that lost its declaration both
+    /// fail closed.
+    #[tokio::test]
+    async fn test_a_broken_declaration_is_refused_before_refresh() {
+        let conn = connect("memory://").execute().await.unwrap();
+        let view = refreshed_computed_view(&conn).await;
+        let native = view.table().as_native().unwrap();
+        let mut dataset = native.dataset.get().await.unwrap().as_ref().clone();
+        dataset
+            .update_schema_metadata(vec![(
+                crate::table::computed_columns::FUNCTION_BINDINGS_META_KEY.to_string(),
+                None,
+            )])
+            .await
+            .unwrap();
+        let err = view.refresh().execute().await.unwrap_err().to_string();
+        assert!(err.contains("references missing binding 'fb_1'"), "{err}");
+
+        let conn = connect("memory://").execute().await.unwrap();
+        let view = refreshed_computed_view(&conn).await;
+        let native = view.table().as_native().unwrap();
+        let mut dataset = native.dataset.get().await.unwrap().as_ref().clone();
+        dataset
+            .replace_field_metadata(vec![(
+                dataset.schema().field("emb").unwrap().id as u32,
+                HashMap::new(),
+            )])
+            .await
+            .unwrap();
+        let err = view.refresh().execute().await.unwrap_err().to_string();
+        assert!(err.contains("does not match binding 'fb_1'"), "{err}");
+    }
+
+    /// An input the view does not project is materialized on every refresh
+    /// path, before the provenance column, with the source's values.
+    #[tokio::test]
+    async fn test_internal_inputs_are_materialized_and_refreshed() {
+        use crate::materialized_view::tests::{computed_field, strict_people, test_binding};
+        let conn = connect("memory://").execute().await.unwrap();
+        let source = strict_people(&conn).await;
+        let mut prepared = crate::materialized_view::prepare_declaration(
+            &source,
+            Some(&[("id".to_string(), "id".to_string())]),
+            None,
+            None,
+        )
+        .await
+        .unwrap();
+        let input = prepared.input_column("name").unwrap();
+        let view = prepared
+            .with_computed_columns(
+                vec![(1, computed_field("emb", "fb_1", &input))],
+                &[test_binding("fb_1", &input, "emb")],
+            )
+            .unwrap()
+            .create("v")
+            .await
+            .unwrap();
+        let names: Vec<String> = view
+            .table()
+            .schema()
+            .await
+            .unwrap()
+            .fields()
+            .iter()
+            .map(|f| f.name().clone())
+            .collect();
+        assert_eq!(names, ["id", "emb", "__input_name", SOURCE_ROW_ID_COLUMN]);
+
+        let unfilled_inputs = || async {
+            view.table()
+                .count_rows(Some("__input_name IS NULL".to_string()))
+                .await
+                .unwrap()
+        };
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::Rebuild
+        );
+        assert_eq!(view.table().count_rows(None).await.unwrap(), 3);
+        assert_eq!(unfilled_inputs().await, 0);
+
+        let more = arrow_array::RecordBatch::try_new(
+            source.schema().await.unwrap(),
+            vec![
+                Arc::new(Int32Array::from(vec![4])),
+                Arc::new(arrow_array::StringArray::from(vec!["d"])),
+            ],
+        )
+        .unwrap();
+        source.add(more).execute().await.unwrap();
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::Incremental
+        );
+        assert_eq!(unfilled_inputs().await, 0);
+        assert_eq!(
+            view.table()
+                .count_rows(Some("__input_name = 'd'".to_string()))
+                .await
+                .unwrap(),
+            1
+        );
+
+        source
+            .update()
+            .column("name", "'z'")
+            .only_if("id = 1")
+            .execute()
+            .await
+            .unwrap();
+        view.refresh().execute().await.unwrap();
+        assert_eq!(
+            view.table()
+                .count_rows(Some("__input_name = 'z'".to_string()))
+                .await
+                .unwrap(),
+            1
+        );
+        assert_eq!(
+            unfilled(&view).await,
+            4,
+            "rewritten and new rows are unfilled"
+        );
+    }
+
+    /// A SQL declaration is filled by `refresh_column` on the view, which
+    /// commits a data replacement and then its freshness stamp; the next
+    /// refresh continues from its watermark and keeps what the fill wrote,
+    /// and only rows the view added since come back unfilled.
+    #[tokio::test]
+    async fn test_a_sql_fill_is_not_drift() {
+        use crate::materialized_view::tests::{people, sql_field};
+        let conn = connect("memory://").execute().await.unwrap();
+        let source = people(&conn).await;
+        let view = crate::materialized_view::prepare_declaration(
+            &source,
+            Some(&[("id".to_string(), "id".to_string())]),
+            None,
+            None,
+        )
+        .await
+        .unwrap()
+        .with_computed_columns(
+            vec![(
+                1,
+                sql_field("next", arrow_schema::DataType::Int32, "id + 1", r#"["id"]"#),
+            )],
+            &[],
+        )
+        .unwrap()
+        .create("v")
+        .await
+        .unwrap();
+        let filled = || async {
+            view.table()
+                .count_rows(Some("next = id + 1".to_string()))
+                .await
+                .unwrap()
+        };
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::Rebuild
+        );
+        assert_eq!(
+            view.table()
+                .refresh_column("next")
+                .await
+                .unwrap()
+                .rows_filled,
+            3
+        );
+        assert_eq!(filled().await, 3);
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::NoOp
+        );
+        assert_eq!(filled().await, 3);
+
+        append_people(&conn, vec![4], vec!["d"]).await;
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::Incremental
+        );
+        assert_eq!(filled().await, 3);
+        assert_eq!(
+            view.table()
+                .refresh_column("next")
+                .await
+                .unwrap()
+                .rows_filled,
+            1
+        );
+        assert_eq!(filled().await, 4);
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::NoOp
+        );
+    }
+
+    /// A fill of a nested computed column writes its child fields; that is
+    /// still a fill, not drift.
+    #[tokio::test]
+    async fn test_a_nested_sql_fill_is_not_drift() {
+        use crate::materialized_view::tests::{people, sql_field};
+        let conn = connect("memory://").execute().await.unwrap();
+        let source = people(&conn).await;
+        let payload = sql_field(
+            "payload",
+            arrow_schema::DataType::Struct(
+                vec![arrow_schema::Field::new(
+                    "value",
+                    arrow_schema::DataType::Utf8,
+                    true,
+                )]
+                .into(),
+            ),
+            "named_struct('value', name)",
+            r#"["name"]"#,
+        );
+        let view = crate::materialized_view::prepare_declaration(
+            &source,
+            Some(&[("name".to_string(), "name".to_string())]),
+            None,
+            None,
+        )
+        .await
+        .unwrap()
+        .with_computed_columns(vec![(1, payload)], &[])
+        .unwrap()
+        .create("v")
+        .await
+        .unwrap();
+        view.refresh().execute().await.unwrap();
+        assert_eq!(
+            view.table()
+                .refresh_column("payload")
+                .await
+                .unwrap()
+                .rows_filled,
+            3
+        );
+        assert_eq!(
+            view.refresh().execute().await.unwrap().mode,
+            RefreshMode::NoOp
+        );
+        assert_eq!(
+            view.table()
+                .count_rows(Some("payload.value = name".to_string()))
+                .await
+                .unwrap(),
+            3
+        );
+    }
 }
@@ -75,6 +75,7 @@ mod create_index;
 pub mod datafusion;
 pub(crate) mod dataset;
 pub mod delete;
+pub mod freshness;
 pub mod lsm_stats;
 pub mod merge;
 pub mod optimize;
@@ -778,7 +779,8 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
            message: "Function columns are supported only on LanceDB Cloud and Enterprise".into(),
        })
    }
-    /// Fill a computed column's unfilled rows.
+    /// Fill a computed column's unfilled rows and recompute those whose
+    /// inputs changed.
    ///
    /// The default returns `NotSupported`; Lance-backed tables override it.
    async fn refresh_column(&self, _column: &str) -> Result<RefreshColumnResult> {
@@ -786,8 +788,8 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
            message: "computed columns are supported only on local tables".into(),
        })
    }
-    /// Fill a computed column's unfilled rows, returning a [`Job`] tracking
-    /// the operation.
+    /// Fill a computed column's unfilled rows and recompute those whose
+    /// inputs changed, returning a [`Job`] tracking the operation.
    async fn refresh_column_async(
        &self,
        _column: &str,
@@ -1749,9 +1751,10 @@ impl Table {
    /// Declared with
    /// [`AddColumnsBuilder::computed`](add_columns::AddColumnsBuilder::computed),
    /// a column starts empty and gets its values here. Fragments appended
-    /// since the last refresh are filled by the next one; fragments already
-    /// filled are left as they are, so the call is idempotent and does not
-    /// observe a mutated input.
+    /// since the last refresh are filled by the next one, and fragments whose
+    /// inputs changed since they were computed are recomputed (see
+    /// [`freshness`](crate::table::freshness)); everything else is left as
+    /// it is.
    ///
    /// Local tables only: a remote refresh runs as a server job, through
    /// [`Table::refresh_column_async`].
@@ -2804,7 +2807,7 @@ impl NativeTable {
        namespace_client: Option<Arc<dyn LanceNamespace>>,
        pushdown_operations: HashSet<NamespaceClientPushdownOperation>,
    ) -> Result<Self> {
-        computed_columns::ensure_no_foreign_declarations(batches.arrow_schema().fields())?;
+        let batches = computed_columns::admit_create_source(batches)?;
        // Default params uses format v1.
        let params = params.unwrap_or(WriteParams {
            ..Default::default()
@@ -2904,6 +2907,7 @@ impl NativeTable {
        pushdown_operations: HashSet<NamespaceClientPushdownOperation>,
        session: Option<Arc<lance::session::Session>>,
    ) -> Result<Self> {
+        let batches = computed_columns::admit_create_source(batches)?;
        // Build table_id from namespace + name for the storage options provider
        let mut table_id = namespace.clone();
        table_id.push(name.to_string());
@@ -5677,7 +5681,7 @@ mod tests {
            TableStatistics {
                num_rows: 250,
                num_indices: 0,
-                total_bytes: 8925,
+                total_bytes: 8969,
                fragment_stats: FragmentStatistics {
                    num_fragments: 11,
                    num_small_fragments: 11,
@@ -60,10 +60,11 @@ impl AddColumnsBuilder {
    /// every fragment that has none -- including fragments appended since the
    /// last refresh.
    ///
-    /// Refresh does not revisit a fragment it has filled, so mutating an input
-    /// leaves the value computed at fill time; recomputing means dropping the
-    /// column and declaring it again. An input cannot be renamed, retyped or
-    /// dropped while a declaration reads it, since the expression names it.
+    /// A refresh also recomputes the rows of a fragment whose inputs changed
+    /// since it was computed (see [`freshness`](super::freshness)), so a
+    /// mutated input is reflected by the next refresh. An input cannot be
+    /// renamed, retyped or dropped while a declaration reads it, since the
+    /// expression names it.
    ///
    /// On LanceDB Cloud and Enterprise the expression is planned by the
    /// server, and the refresh runs as a server job -- see
@@ -21,6 +21,7 @@
 //! [`computed_columns`] and [`computed_column_from_field`] read declarations
 //! back off a schema.

+use futures::StreamExt;
 use std::collections::{BTreeSet, HashMap, HashSet};
 use std::sync::Arc;

@@ -70,6 +71,22 @@ pub const FUNCTION_BINDINGS_META_KEY: &str = "lancedb::function_bindings";
 /// Version of the schema-level Function binding envelope.
 pub const FUNCTION_BINDINGS_VERSION: u32 = 1;

+/// Field metadata key holding `{fragment id -> input signature}` as JSON,
+/// recorded by the refresh that last computed each fragment. Outside the
+/// declaration namespace on purpose: a declaration is immutable through
+/// metadata edits, this is rewritten by every refresh. Seeded empty at
+/// declaration, so a column is tracked from birth; a column without it was
+/// declared before signatures existed.
+pub const SOURCE_SIGNATURE_META_KEY: &str = "computed_refresh.source_signature";
+
+/// Field metadata key holding the definition digest a column was last
+/// computed under. A change to it makes every row stale.
+pub const DEFINITION_VERSION_META_KEY: &str = "computed_refresh.definition_version";
+
+/// Field metadata key holding the table version the signature map describes:
+/// where a refresh starts following compactions to carry freshness forward.
+pub const RECORDED_AT_VERSION_META_KEY: &str = "computed_refresh.recorded_at_version";
+
 /// Value of [`KIND_META_KEY`] for a column defined by a SQL expression.
 pub const SQL_KIND: &str = "sql";

@@ -138,6 +155,7 @@ fn computed_column_metadata(expression: &str, inputs: &[String]) -> HashMap<Stri
            INPUTS_META_KEY.to_string(),
            serde_json::to_string(inputs).unwrap_or_else(|_| "[]".to_string()),
        ),
+        (SOURCE_SIGNATURE_META_KEY.to_string(), "{}".to_string()),
    ])
 }

@@ -162,6 +180,7 @@ pub fn function_computed_column_metadata(
            INPUTS_META_KEY.to_string(),
            serde_json::to_string(inputs).unwrap_or_else(|_| "[]".to_string()),
        ),
+        (SOURCE_SIGNATURE_META_KEY.to_string(), "{}".to_string()),
    ])
 }

@@ -1294,7 +1313,8 @@ pub(crate) fn ensure_not_an_input(schema: &SchemaRef, paths: &[&str]) -> Result<
 }

 /// Reject a write that supplies values for a computed column directly:
-/// only refresh materializes one, and refresh never revisits a filled row.
+/// only refresh materializes one, and only refresh decides what it
+/// recomputes.
 pub(crate) fn ensure_not_written<'a>(
    schema: &ArrowSchema,
    written: impl IntoIterator<Item = &'a str>,
@@ -1338,6 +1358,106 @@ pub(crate) fn ensure_batch_writes_no_computed_values(
    Ok(())
 }

+/// Validate every computed-column declaration `schema` carries against the
+/// schema itself: every field with declaration metadata is a complete
+/// declaration, a SQL declaration re-plans to the field it declares, a
+/// Function declaration satisfies the binding contract, and no declaration
+/// reads another computed column. What passes here is what `refresh_column`
+/// can execute.
+pub(crate) fn ensure_declarations_are_planned(schema: &ArrowSchema) -> Result<()> {
+    let invalid = |message: String| Error::InvalidInput { message };
+    // A field with any declaration key is a declaration; a partial one is
+    // not "no declaration", it is a broken one.
+    for field in schema.fields() {
+        if field.metadata().keys().any(|k| is_declaration_key(k))
+            && computed_column_from_field(field).is_none()
+        {
+            return Err(invalid(format!(
+                "field '{}' carries an incomplete computed-column declaration",
+                field.name()
+            )));
+        }
+    }
+    let declared: HashSet<String> = computed_columns(schema)
+        .into_iter()
+        .map(|c| c.name)
+        .collect();
+    for column in computed_columns(schema) {
+        let field = schema.field_with_name(&column.name)?;
+        if !field.is_nullable() {
+            return Err(invalid(format!(
+                "computed column '{}' must be nullable until a refresh fills it",
+                column.name
+            )));
+        }
+        match &column.kind {
+            ComputedColumnKind::Sql { expression } => {
+                let others: Vec<ArrowField> = schema
+                    .fields()
+                    .iter()
+                    .filter(|f| f.name() != &column.name)
+                    .map(|f| f.as_ref().clone())
+                    .collect();
+                let bound = bind(Arc::new(ArrowSchema::new(others)), &column.name, expression)?;
+                if let Some(input) = bound.roots.iter().find(|r| declared.contains(*r)) {
+                    return Err(invalid(format!(
+                        "computed column '{}' reads computed column '{input}'",
+                        column.name
+                    )));
+                }
+                if &bound.data_type != field.data_type() {
+                    return Err(invalid(format!(
+                        "computed column '{}' is declared as {} but its expression yields {}",
+                        column.name,
+                        field.data_type(),
+                        bound.data_type
+                    )));
+                }
+                let mut declared_inputs = column.inputs.clone();
+                declared_inputs.sort();
+                if declared_inputs != bound.inputs {
+                    return Err(invalid(format!(
+                        "computed column '{}' declares inputs {:?} but its expression reads {:?}",
+                        column.name, declared_inputs, bound.inputs
+                    )));
+                }
+            }
+            ComputedColumnKind::Function { binding_id, .. } => {
+                // The binding validator resolves each input's leaf; the
+                // no-computed-input rule is about the root it hangs from.
+                let bindings = function_bindings(schema)?;
+                let Some(binding) = bindings.iter().find(|b| b.binding_id() == binding_id) else {
+                    continue; // reported by the binding validator below
+                };
+                // Roots come from the canonical path parser: a quoted
+                // top-level name may itself contain a dot.
+                if let Some(input) = binding
+                    .inputs()
+                    .iter()
+                    .filter_map(|input| resolve_field_path(schema, &input.field_path).ok())
+                    .map(|resolved| resolved.root.name().as_str())
+                    .find(|r| declared.contains(*r))
+                {
+                    return Err(invalid(format!(
+                        "computed column '{}' reads computed column '{input}'",
+                        column.name
+                    )));
+                }
+            }
+            ComputedColumnKind::Unrecognized { kind } => {
+                return Err(Error::NotSupported {
+                    message: format!(
+                        "computed column '{}' is defined by '{kind}', which this version \
+                         of lancedb cannot fill",
+                        column.name
+                    ),
+                });
+            }
+        }
+    }
+    ensure_supported_function_metadata(schema)
+}
+
 /// Reject fields carrying declaration metadata that did not come through
 /// [`plan`]. One authority for creation, overwrite and raw transforms.
 pub(crate) fn ensure_no_foreign_declarations<'a>(
@@ -1369,7 +1489,9 @@ fn ensure_no_foreign_declaration(field: &ArrowField) -> Result<()> {
 /// kind, the expression, the inputs -- would bypass that validation or move
 /// a binding out from under a refresh. Drop the column and declare it again.
 pub(crate) fn is_declaration_key(key: &str) -> bool {
-    key == COMPUTED_COLUMN_META_KEY || key.starts_with("computed_column.")
+    key == COMPUTED_COLUMN_META_KEY
+        || key.starts_with("computed_column.")
+        || key.starts_with("computed_refresh.")
 }

 /// Reject retyping a computed column itself.
@@ -1796,6 +1918,54 @@ pub(super) async fn add_foreign_kind(table: &crate::Table, name: &str, kind: &st
    .unwrap();
 }

+/// Admit a table's initial data: every declaration it carries is validated,
+/// and the stream refuses any batch with values in a computed column, whose
+/// values come from refresh alone. One boundary for every way a table is
+/// created.
+pub(crate) fn admit_create_source<S: lance_datafusion::utils::StreamingWriteSource>(
+    batches: S,
+) -> Result<UnfilledDeclarations<S>> {
+    let schema = batches.arrow_schema();
+    ensure_declarations_are_planned(&schema)?;
+    let declared = computed_columns(&schema)
+        .into_iter()
+        .map(|c| c.name)
+        .collect();
+    Ok(UnfilledDeclarations {
+        inner: batches,
+        declared,
+    })
+}
+
+/// A write source whose computed columns must arrive unfilled.
+pub(crate) struct UnfilledDeclarations<S> {
+    inner: S,
+    declared: Vec<String>,
+}
+
+impl<S: lance_datafusion::utils::StreamingWriteSource> lance_datafusion::utils::StreamingWriteSource
+    for UnfilledDeclarations<S>
+{
+    fn arrow_schema(&self) -> SchemaRef {
+        self.inner.arrow_schema()
+    }
+
+    fn into_stream(self) -> datafusion_physical_plan::SendableRecordBatchStream {
+        if self.declared.is_empty() {
+            return self.inner.into_stream();
+        }
+        let schema = self.inner.arrow_schema();
+        let declared = self.declared;
+        let stream = self.inner.into_stream().map(move |batch| {
+            let batch = batch?;
+            ensure_batch_writes_no_computed_values(&declared, &batch)
+                .map_err(|e| datafusion_common::DataFusionError::External(Box::new(e)))?;
+            Ok(batch)
+        });
+        Box::pin(datafusion_physical_plan::stream::RecordBatchStreamAdapter::new(schema, stream))
+    }
+}
+
 #[cfg(test)]
 mod tests {
    /// The gate's reproducer: the validator applies the same schema-level
@@ -2646,6 +2816,8 @@ mod tests {
        );
    }

+    /// A create carries a declaration only if it re-plans completely; this
+    /// one lacks its inputs and is refused before its forged value matters.
    #[tokio::test]
    async fn test_create_table_cannot_inject_a_declaration() {
        let conn = connect("memory://").execute().await.unwrap();
@@ -2673,7 +2845,7 @@ mod tests {
            .await
            .unwrap_err();
        assert!(
-            matches!(&err, Error::InvalidInput { message } if message.contains("computed()")),
+            matches!(&err, Error::InvalidInput { message } if message.contains("computed column 'doubled'")),
            "{err:?}"
        );
    }
@@ -3,9 +3,9 @@

 //! Filling computed columns.
 //!
-//! A row without a value gets one; a row that has one keeps it. Refresh is
-//! therefore idempotent and does not observe input mutation -- once a row is
-//! filled, changing what the expression reads leaves the stored result alone.
+//! A row without a value gets one; a row that has one keeps it unless its
+//! fragment's inputs moved since it was computed, which `freshness` decides
+//! from the manifest and stamps after every fill.
 //!
 //! A column's computed inputs are filled first -- the dependency graph is
 //! walked once, each reachable column filled once in dependency order, each
@@ -31,6 +31,7 @@

 use std::collections::HashSet;
 use std::sync::Arc;
+use std::sync::atomic::{AtomicU64, Ordering};

 use arrow_array::{
    Array, ArrayRef, BooleanArray, LargeBinaryArray, RecordBatch, RecordBatchOptions, StructArray,
@@ -48,6 +49,7 @@ use lance_core::datatypes::{BlobHandling, Schema as LanceSchema};
 use serde::{Deserialize, Serialize};

 use super::computed_columns::{BoundExpression, ComputedColumnKind, computed_column_from_field};
+use super::freshness::{self, SignatureMap, StalenessPlan};
 use super::{BaseTable, NativeTable};
 use crate::job::Job;
 use crate::{Error, Result};
@@ -110,28 +112,81 @@ async fn execute_refresh_column_with_source(
    };
    let output_is_blob = field.is_blob_v2();

+    // Which fragments the null filter cannot speak for: their inputs moved
+    // since they were computed, or the definition did. Decided once, from
+    // the manifest the values are read from.
+    let inputs = freshness::fields_for_paths(dataset.schema(), &bound.inputs)?;
+    let definition = freshness::definition_version(&expression);
+    let staleness = freshness::staleness_against(&dataset, column, &definition, &inputs).await?;
+
    let mut rows_filled = 0u64;
    let mut replacements = Vec::new();
+    // Fragments this refresh computed in full, signed at the version read.
+    let mut computed = SignatureMap::new();
    for fragment in dataset.get_fragments() {
-        let gained = count_fragment_gains(&dataset, &fragment, &bound, column).await?;
-        if gained == 0 {
-            continue;
+        let fragment_id = u32::try_from(fragment.id()).map_err(|_| Error::Runtime {
+            message: format!("fragment id {} does not fit a signature map", fragment.id()),
+        })?;
+        // A recompute rewrites every live row, so it is staged without the
+        // probe and counted as it fills; a null fill probes first, since a
+        // fragment with nothing to gain is not worth a write.
+        let recompute = staleness.is_dirty(fragment_id);
+        let whole = recompute || {
+            let (gained, unfilled) =
+                count_fragment_gains(&dataset, &fragment, &bound, column).await?;
+            if gained == 0 {
+                continue;
+            }
+            rows_filled += gained;
+            unfilled == u64::try_from(fragment.count_rows(None).await?).unwrap_or(u64::MAX)
+        };
+        if whole {
+            computed.insert(
+                fragment_id,
+                freshness::fragment_input_signature(fragment.metadata(), &inputs)?,
+            );
        }
-        rows_filled += gained;
-        let values =
-            fill_stream(&dataset, &fragment, bound.clone(), column, output_is_blob).await?;
+        let gained = Arc::new(AtomicU64::new(0));
+        let values = fill_stream(
+            &dataset,
+            &fragment,
+            bound.clone(),
+            column,
+            output_is_blob,
+            recompute,
+            gained.clone(),
+        )
+        .await?;
        replacements.push(fragment.write_columns(values, &column_schema).await?);
+        if recompute {
+            rows_filled += gained.load(Ordering::Relaxed);
+        }
    }

    let source_version = dataset.version().version;
    if replacements.is_empty() {
+        // Nothing to fill; the stamp may still have something to record -- a
+        // column not yet enrolled, or fragments a compaction carried.
+        let mut latest = (*dataset).clone();
+        let stamped = record(
+            &mut latest,
+            (&dataset, &staleness),
+            column,
+            &definition,
+            &inputs,
+            computed,
+        )
+        .await;
+        if stamped.is_some() {
+            table.dataset.update(latest);
+        }
        return Ok(RefreshExecution {
            result: RefreshColumnResult {
                rows_filled: 0,
-                version: source_version,
+                version: stamped.unwrap_or(source_version),
            },
            source_version,
-            published_version: None,
+            published_version: stamped,
        });
    }

@@ -149,7 +204,17 @@ async fn execute_refresh_column_with_source(
    )
    .await?;

-    let version = new_dataset.version().version;
+    let mut new_dataset = new_dataset;
+    let version = record(
+        &mut new_dataset,
+        (&dataset, &staleness),
+        column,
+        &definition,
+        &inputs,
+        computed,
+    )
+    .await
+    .unwrap_or(new_dataset.version().version);
    table.dataset.update(new_dataset);
    Ok(RefreshExecution {
        result: RefreshColumnResult {
@@ -161,6 +226,31 @@ async fn execute_refresh_column_with_source(
    })
 }

+/// Stamp the input state the refresh computed from (see
+/// [`freshness::record_freshness`]); the version the stamp landed at, which
+/// is the last one the refresh wrote. Never fails the refresh: the values
+/// are committed, and a missing stamp only costs a recompute next time.
+async fn record(
+    latest: &mut Dataset,
+    pinned: (&Dataset, &StalenessPlan),
+    column: &str,
+    definition: &str,
+    inputs: &freshness::InputFields,
+    computed: SignatureMap,
+) -> Option<u64> {
+    match freshness::record_freshness(latest, Some(pinned), column, definition, inputs, computed)
+        .await
+    {
+        Ok(record) => record.version,
+        Err(error) => {
+            log::warn!(
+                "could not record the input state computed column '{column}' was refreshed from ({error}); its fragments will recompute on the next refresh"
+            );
+            None
+        }
+    }
+}
+
 /// Refuse while a computed input still has rows a refresh of it would fill:
 /// read now, its placeholder null would be evaluated as a value and kept.
 async fn ensure_inputs_filled(
@@ -188,7 +278,9 @@ async fn ensure_inputs_filled(
        let input_bound = super::computed_columns::bind(schema.clone(), input, expression)?;
        let mut unfilled = 0u64;
        for fragment in dataset.get_fragments() {
-            unfilled += count_fragment_gains(dataset, &fragment, &input_bound, input).await?;
+            unfilled += count_fragment_gains(dataset, &fragment, &input_bound, input)
+                .await?
+                .0;
        }
        if unfilled > 0 {
            return Err(Error::InvalidInput {
@@ -436,32 +528,35 @@ fn blob_array_from_binary(
 /// Scans only the unfilled live rows -- deleted rows never reach the
 /// expression here, the filter having already excluded them -- and counts the
 /// non-null results. Exact, so it is both the staging decision and the
-/// fragment's contribution to `rows_filled`.
+/// fragment's contribution to `rows_filled`. Returns the gains and the rows
+/// scanned.
 async fn count_fragment_gains(
    dataset: &Dataset,
    fragment: &FileFragment,
    bound: &BoundExpression,
    column: &str,
-) -> Result<u64> {
+) -> Result<(u64, u64)> {
    let mut scanner = dataset.scan();
    scanner
        .with_fragments(vec![fragment.metadata().clone()])
        .with_row_id()
-        .filter(&format!("{} IS NULL", quote_identifier(column)))?
-        .project(&bound.roots)?;
+        .project(&bound.roots)?
+        .filter(&format!("{} IS NULL", quote_identifier(column)))?;
    configure_blob_inputs(&mut scanner, dataset.schema(), bound, None)?;

    let mut gained = 0u64;
+    let mut considered = 0u64;
    let mut batches = scanner.try_into_stream().await?;
    while let Some(batch) = batches.try_next().await? {
        let evaluated = evaluate(bound, &evaluation_batch(&batch, bound, None)?)?;
        gained += (batch.num_rows() - evaluated.null_count()) as u64;
+        considered += batch.num_rows() as u64;
    }
-    Ok(gained)
+    Ok((gained, considered))
 }

 /// Stream one fragment's column in physical order, filling the unfilled live
-/// rows and keeping every other value.
+/// rows -- every live row, for a recompute -- and keeping every other value.
 ///
 /// Deleted rows are carried through so the values line up positionally with
 /// the fragment's data files; they are never read back, but the column file
@@ -472,6 +567,8 @@ async fn fill_stream(
    bound: Arc<BoundExpression>,
    column: &str,
    output_is_blob: bool,
+    recompute: bool,
+    gained: Arc<AtomicU64>,
 ) -> Result<impl Stream<Item = lance_core::Result<RecordBatch>> + Send + use<>> {
    let mut projection: Vec<String> = bound.roots.clone();
    projection.push(column.to_string());
@@ -521,14 +618,20 @@ async fn fill_stream(
            .column_by_name(ROW_ID)
            .ok_or_else(|| missing(ROW_ID))?;

-        // Only an unfilled live row gains a value; a deleted row has a null
-        // row id and keeps its (null) slot.
-        let unfilled = arrow::compute::is_null(existing.as_ref())?;
+        // Only an unfilled live row gains a value, or every live row under a
+        // recompute; a deleted row has a null row id and keeps its (null) slot.
        let live = arrow::compute::is_not_null(row_ids.as_ref())?;
-        let fill = arrow::compute::and(&unfilled, &live)?;
+        let fill = if recompute {
+            live
+        } else {
+            let unfilled = arrow::compute::is_null(existing.as_ref())?;
+            arrow::compute::and(&unfilled, &live)?
+        };
        let keep = arrow::compute::not(&fill)?;

        let computed = evaluate(&bound, &evaluation_batch(&batch, &bound, Some(&keep))?)?;
+        let values = arrow::compute::and(&fill, &arrow::compute::is_not_null(&computed)?)?;
+        gained.fetch_add(values.true_count() as u64, Ordering::Relaxed);
        let merged = arrow_select::zip::zip(&fill, &computed, existing)?;
        let merged = if output_is_blob {
            blob_array_from_binary(&merged, projected.field(0))?
@@ -737,7 +840,8 @@ mod tests {
            .await
            .unwrap();
        assert_eq!(no_op.rows_assigned, 0);
-        assert_eq!(no_op.source_version, 3);
+        // The fill, then the stamp recording what it computed from.
+        assert_eq!(no_op.source_version, 4);
        assert_eq!(no_op.published_version, None);
    }

@@ -777,8 +881,8 @@ mod tests {
    }

    /// A row is filled only by gaining a value, so an expression yielding null
-    /// settles at once instead of re-selecting the same rows forever. Nothing
-    /// is staged, so the version does not move either.
+    /// settles at once instead of re-selecting the same rows forever: the
+    /// second refresh finds the fragment signed and moves nothing.
    #[tokio::test]
    async fn test_refresh_converges_on_a_null_result() {
        let table = table_with("refresh_null_result", vec![1, 2, 3]).await;
@@ -792,28 +896,145 @@ mod tests {

        let first = table.refresh_column("maybe").await.unwrap();
        assert_eq!(first.rows_filled, 0);
-        assert_eq!(first.version, declared);
+        assert!(first.version > declared);
        assert_eq!(read(&table, "maybe").await, vec![None, None, None]);

        let again = table.refresh_column("maybe").await.unwrap();
        assert_eq!(again.rows_filled, 0);
-        assert_eq!(again.version, declared);
+        assert_eq!(again.version, first.version);
    }

-    /// The contract's boundary: a filled fragment is not revisited, so
-    /// mutating an input leaves the value computed at fill time.
+    /// A filled row whose input moved is recomputed: the update rewrites
+    /// the row into a fragment the stamp never signed, and only that one.
    #[tokio::test]
-    async fn test_refresh_does_not_observe_input_mutation() {
-        let table = table_with("refresh_mutation", vec![1]).await;
+    async fn test_refresh_recomputes_a_row_whose_input_moved() {
+        let table = table_with("refresh_mutation", vec![1, 2]).await;
        declare_doubled(&table).await.unwrap();
+        append(&table, vec![5]).await;
        table.refresh_column("doubled").await.unwrap();
-        assert_eq!(read(&table, "doubled").await, vec![Some(2)]);
+        assert_eq!(
+            read(&table, "doubled").await,
+            vec![Some(2), Some(4), Some(10)]
+        );

-        table.update().column("x", "3").execute().await.unwrap();
+        table
+            .update()
+            .column("x", "7")
+            .only_if("x = 5")
+            .execute()
+            .await
+            .unwrap();
+
+        let again = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(again.rows_filled, 1);
+        assert_eq!(
+            read(&table, "doubled").await,
+            vec![Some(2), Some(4), Some(14)]
+        );
+
+        let settled = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(settled.rows_filled, 0);
+        assert_eq!(settled.version, again.version);
+    }
+
+    /// A deleted row is never computed and the rows that stay keep their
+    /// values: a delete recomputes nothing and stamps nothing.
+    #[tokio::test]
+    async fn test_a_delete_recomputes_nothing() {
+        let table = table_with("refresh_delete", vec![1, 2, 3]).await;
+        declare_doubled(&table).await.unwrap();
+        let filled = table.refresh_column("doubled").await.unwrap();
+
+        table.delete("x = 2").await.unwrap();
+        let deleted = table.version().await.unwrap();

        let again = table.refresh_column("doubled").await.unwrap();
        assert_eq!(again.rows_filled, 0);
-        assert_eq!(read(&table, "doubled").await, vec![Some(2)]);
+        assert_eq!(again.version, deleted);
+        assert!(deleted > filled.version);
+        assert_eq!(read(&table, "doubled").await, vec![Some(2), Some(6)]);
+    }
+
+    /// Compaction copies inputs unchanged, so a fragment it builds from
+    /// signed ones is fresh: the refresh recomputes nothing and only records
+    /// the new fragment.
+    #[tokio::test]
+    async fn test_a_compaction_of_signed_fragments_recomputes_nothing() {
+        let table = table_with("refresh_compact_signed", vec![1, 2]).await;
+        declare_doubled(&table).await.unwrap();
+        append(&table, vec![5]).await;
+        table.refresh_column("doubled").await.unwrap();
+
+        table
+            .optimize(crate::table::OptimizeAction::Compact {
+                options: crate::table::CompactionOptions::default(),
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+        let compacted = table.version().await.unwrap();
+
+        let carried = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(carried.rows_filled, 0);
+        assert_eq!(carried.version, compacted + 1);
+        let settled = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(settled.version, carried.version);
+        assert_eq!(
+            read(&table, "doubled").await,
+            vec![Some(2), Some(4), Some(10)]
+        );
+    }
+
+    /// A column declared before signatures existed has no map. Its first
+    /// refresh keeps the null-fill contract and enrolls what it read from;
+    /// from then on a moved input is recomputed like any other.
+    #[tokio::test]
+    async fn test_an_unsigned_column_is_enrolled_by_its_first_refresh() {
+        let table = table_with("refresh_legacy", vec![1, 2]).await;
+        declare_doubled(&table).await.unwrap();
+        table.refresh_column("doubled").await.unwrap();
+        table
+            .update()
+            .column("x", "3")
+            .only_if("x = 1")
+            .execute()
+            .await
+            .unwrap();
+
+        let native = table.as_native().unwrap();
+        let mut dataset = (*native.dataset.get().await.unwrap()).clone();
+        let declaration = dataset
+            .schema()
+            .field("doubled")
+            .unwrap()
+            .metadata
+            .iter()
+            .filter(|(key, _)| !key.starts_with("computed_refresh."))
+            .map(|(key, value)| (key.clone(), value.clone()))
+            .collect::<Vec<_>>();
+        dataset
+            .update_field_metadata()
+            .replace("doubled", declaration)
+            .unwrap()
+            .await
+            .unwrap();
+        table.checkout_latest().await.unwrap();
+
+        // Null-fill only: the moved row keeps the value it was filled with.
+        let enrolled = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(enrolled.rows_filled, 0);
+        assert_eq!(read(&table, "doubled").await, vec![Some(2), Some(4)]);
+
+        table
+            .update()
+            .column("x", "5")
+            .only_if("x = 3")
+            .execute()
+            .await
+            .unwrap();
+        let again = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(again.rows_filled, 1);
+        assert_eq!(read(&table, "doubled").await, vec![Some(4), Some(10)]);
    }

    /// A row rewrite before the first refresh materializes the declared
@@ -831,11 +1052,11 @@ mod tests {
        assert_eq!(read(&table, "doubled").await, vec![Some(6)]);
    }

-    /// The contract holds row by row, not fragment by fragment: revisiting a
-    /// fragment to fill one row must not recompute a filled row sitting beside
-    /// it, even where the input behind it has since changed.
+    /// A fragment compacted out of one the stamp never signed cannot vouch
+    /// for any of its rows: every live row is recomputed, the moved one
+    /// included.
    #[tokio::test]
-    async fn test_refresh_does_not_recompute_a_filled_row_beside_an_unfilled_one() {
+    async fn test_a_compaction_of_an_unsigned_fragment_recomputes_it() {
        let table = table_with("refresh_mixed", vec![1, 2]).await;
        declare_doubled(&table).await.unwrap();
        table.refresh_column("doubled").await.unwrap();
@@ -857,18 +1078,70 @@ mod tests {
            .unwrap();

        let result = table.refresh_column("doubled").await.unwrap();
-        assert_eq!(result.rows_filled, 1);
-        // 2 is the mutated row keeping the value it was filled with, not 200.
+        assert_eq!(result.rows_filled, 3);
+        assert_eq!(
+            read(&table, "doubled").await,
+            vec![Some(4), Some(10), Some(200)]
+        );
+    }
+
+    /// The gate's reproducer: a raw lance append may carry a value for the
+    /// computed column. Compaction cannot certify it, so the product is
+    /// recomputed and the supplied value replaced.
+    #[tokio::test]
+    async fn test_raw_append_values_are_not_trusted_after_compaction() {
+        use arrow_array::RecordBatchIterator;
+        use lance::Dataset;
+        use lance::dataset::{WriteMode, WriteParams};
+
+        let dir = tempfile::tempdir().unwrap();
+        let conn = connect(dir.path().to_str().unwrap())
+            .execute()
+            .await
+            .unwrap();
+        let batch = record_batch!(("x", Int32, [1, 2])).unwrap();
+        let table = conn
+            .create_table("raw_append", batch)
+            .execute()
+            .await
+            .unwrap();
+        declare_doubled(&table).await.unwrap();
+        table.refresh_column("doubled").await.unwrap();
+
+        let batch = record_batch!(("x", Int32, [5]), ("doubled", Int32, [Some(999_i32)])).unwrap();
+        let schema = batch.schema();
+        let uri = table.uri().await.unwrap();
+        Dataset::write(
+            RecordBatchIterator::new(vec![Ok(batch)], schema),
+            &uri,
+            Some(WriteParams {
+                mode: WriteMode::Append,
+                ..Default::default()
+            }),
+        )
+        .await
+        .unwrap();
+        table.checkout_latest().await.unwrap();
+        table
+            .optimize(crate::table::OptimizeAction::Compact {
+                options: crate::table::CompactionOptions::default(),
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+
+        let result = table.refresh_column("doubled").await.unwrap();
+        assert_eq!(result.rows_filled, 3);
        assert_eq!(
            read(&table, "doubled").await,
            vec![Some(2), Some(4), Some(10)]
        );
    }

-    /// Filling a fragment must not disturb the values it already holds, which
-    /// is what makes a compaction-mixed fragment safe to revisit.
+    /// An appended fragment holds no values, so compacting it into a signed
+    /// one leaves the product fresh: only the appended rows are filled.
    #[tokio::test]
-    async fn test_refresh_preserves_already_filled_rows() {
+    async fn test_a_compaction_with_an_appended_fragment_fills_only_its_rows() {
        let table = table_with("refresh_preserves", vec![1, 2]).await;
        declare_doubled(&table).await.unwrap();
        table.refresh_column("doubled").await.unwrap();
@@ -995,7 +1268,8 @@ mod tests {
        assert_eq!(result.rows_failed, 0);
        assert_eq!(result.rows_remaining, 0);
        assert_eq!(result.source_version, 2);
-        assert_eq!(result.published_version, Some(3));
+        // The fill lands at 3; the stamp recording its inputs is published at 4.
+        assert_eq!(result.published_version, Some(4));
        assert_eq!(job.status().await.unwrap(), "finished");
        assert_eq!(
            read(&table, "doubled").await,
@@ -1059,31 +1333,32 @@ mod tests {
        assert_eq!(read(&table, "quotient").await, vec![Some(10)]);
    }

-    /// The gate's reproducer: an already-filled row's value must not be
-    /// re-evaluated either -- its input may have mutated into one the
-    /// expression chokes on.
+    /// A filled row whose input moved is re-evaluated, and a row whose
+    /// input did not move is not: the untouched fragment is never read, so
+    /// its poison input is never reached.
    #[tokio::test]
-    async fn test_a_filled_rows_value_is_never_evaluated() {
-        let table = table_with("refresh_filled_poison", vec![1, 2]).await;
+    async fn test_only_a_moved_rows_value_is_re_evaluated() {
+        let table = table_with("refresh_filled_poison", vec![1, 0]).await;
        table
            .add_columns()
-            .computed("quotient", "10 / x")
+            .computed("quotient", "10 / coalesce(nullif(x, 0), 1)")
            .execute()
            .await
            .unwrap();
        table.refresh_column("quotient").await.unwrap();
+        assert_eq!(read(&table, "quotient").await, vec![Some(10), Some(10)]);

+        append(&table, vec![5]).await;
        table
            .update()
-            .column("x", "0")
+            .column("x", "2")
            .only_if("x = 1")
            .execute()
            .await
            .unwrap();
-        append(&table, vec![5]).await;

        let result = table.refresh_column("quotient").await.unwrap();
-        assert_eq!(result.rows_filled, 1);
+        assert_eq!(result.rows_filled, 2);
        assert_eq!(
            read(&table, "quotient").await,
            vec![Some(2), Some(5), Some(10)]
@@ -19,6 +19,7 @@ use lancedb::{
    connect, connect_namespace,
    database::listing::{
        ListingDatabaseOptions, NewTableConfig, OPT_NEW_TABLE_ENABLE_STABLE_ROW_IDS,
+        OPT_NEW_TABLE_STORAGE_VERSION,
    },
    query::{ExecutableQuery, QueryBase},
    table::{AddDataMode, CompactionOptions, OptimizeAction, OptimizeStats, WriteOptions},
@@ -146,7 +147,10 @@ async fn non_blob_table_keeps_default_format_and_row_id_setting() -> Result<()>
    let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int64, false)]));
    let table = db.create_empty_table("t", schema).execute().await?;

-    assert!(!supports_blob_v2(storage_format_version(&table).await));
+    assert_eq!(
+        storage_format_version(&table).await,
+        LanceFileVersion::Stable.resolve()
+    );
    assert!(!uses_stable_row_ids(&table).await);
    Ok(())
 }
@@ -809,7 +813,11 @@ async fn fetch_blobs_rejects_unknown_column() -> Result<()> {
 #[tokio::test]
 async fn fetch_blobs_rejects_legacy_v1_blob_column() -> Result<()> {
    let tmp = tempdir().unwrap();
-    let db = connect(tmp.path().to_str().unwrap()).execute().await?;
+    // Legacy v1 blob columns are only writable at file version <= 2.1.
+    let db = connect(tmp.path().to_str().unwrap())
+        .storage_options([(OPT_NEW_TABLE_STORAGE_VERSION, "2.1")])
+        .execute()
+        .await?;
    let legacy = Field::new("image", DataType::LargeBinary, true).with_metadata(
        std::collections::HashMap::from([("lance-encoding:blob".to_string(), "true".to_string())]),
    );
Author	SHA1	Message	Date
Wyatt Alt	7b8ceb355c	feat: recompute computed column rows whose inputs changed refresh_column fills nulls, so once a row has a value nothing revisits it: an update to one of its inputs, or a definition change, leaves the computed value stale for good. This stamps the column's field metadata with the definition it was computed under and a per-fragment signature of the input storage it was read from (input data files and overlays; not the deletion file, since a delete changes no surviving value). A refresh recomputes every live row of a fragment whose stamp disagrees with the manifest, then records what it computed from in a second commit after the fill. A compacted fragment inherits freshness through the Rewrite lineage when every fragment it was built from was signed, or was appended since the stamp, never had an input moved, and left its rows of the product unfilled (a raw append may supply a value; the product's data is the evidence, and the null fill covers those rows); otherwise it recomputes. A column declared before the stamps existed keeps the null-fill contract on its first refresh, which enrolls it as it stood. The stamp is a metadata-only commit on the computed columns, so a materialized view's drift check treats it like the fill. The core lives in `table::freshness` so a remote refresh can share the contract.	2026-09-11 04:52:16 +00:00
Lance Release	d3077b7641	Bump version: 0.39.0-beta.5 → 0.39.0-beta.6	2026-09-08 12:14:12 +00:00
LanceDB RobotandJack Ye	3e3878b223	chore: update lance dependency to v12.0.0-beta.15 (#4143 ) Updates the Rust workspace Lance dependencies, Cargo lockfile, and Java lance-core from v12.0.0-beta.14 to [v12.0.0-beta.15](https://github.com/lance-format/lance/releases/tag/v12.0.0-beta.15). No compatibility fixes were required; `cargo clippy --quiet --workspace --tests --all-features -- -D warnings`, `cargo fmt --all --quiet`, and `git diff --check` passed. --------- Co-authored-by: Jack Ye <yezhaoqin@gmail.com>	2026-09-08 05:12:45 -07:00
Lance Release	19fb665c76	Bump version: 0.39.0-beta.4 → 0.39.0-beta.5	2026-09-08 12:03:46 +00:00
Wyatt AltandClaude Fable 5.1	1f95398c34	feat: function columns on materialized views (#4119 ) A view could not carry a column it does not compute: the definition planned every output as a SQL expression, and the refresh engine treated any commit it did not make as drift and rebuilt. Servers fill such columns on tables with a separate job, as computed columns bound to a registered function, and want the same column on a view. This lets a declaration add computed columns, placed at their positions in the select list and validated by the existing computed-column contract, with the view created in one commit. Refresh writes those columns NULL on every path and never reads them, so a rewritten row comes back unfilled, and a commit that rewrites only computed columns is recognised as a fill rather than drift, so the next refresh carries on incrementally. A source column a computed column reads without the view projecting it is held as an internal projection, so the select list stays the view's column list. Nothing in the stored definition changes; an older reader fails closed on the schema check. Two smaller changes ride along because the feature needs them: an identity projection keeps its source column's nullability, with the schema check accepting a nullable physical field for a non-null planned one so existing views keep refreshing; and `prepare_declaration` takes `Option` projections, so an empty list declares no projection rather than `SELECT *`. --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-08 04:38:22 -07:00