This is an automated email from the ASF dual-hosted git repository.
laskoviymishka pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/iceberg-go.git
The following commit(s) were added to refs/heads/main by this push:
new 7dd736493 fix(table): validate deletion vector fields in RowDelta
(#2004 follow-up) (#2016)
7dd736493 is described below
commit 7dd7364930b6c888aa939a4fe19038ea51cc1086
Author: Badal Prasad Singh <[email protected]>
AuthorDate: Thu Sep 24 02:47:03 2026 +0530
fix(table): validate deletion vector fields in RowDelta (#2004 follow-up)
(#2016)
* fix: reject dvs below v3
Signed-off-by: badalprasadsingh <[email protected]>
* minor follow ups
Signed-off-by: badalprasadsingh <[email protected]>
* minor
Signed-off-by: badalprasadsingh <[email protected]>
* chore: minor
Signed-off-by: badalprasadsingh <[email protected]>
* minor
Signed-off-by: badalprasadsingh <[email protected]>
* minor
Signed-off-by: badalprasadsingh <[email protected]>
---------
Signed-off-by: badalprasadsingh <[email protected]>
---
table/data_file_meta_test.go | 2 +-
table/row_delta.go | 9 ++-
table/row_delta_test.go | 88 +++++++++++++++++----------
table/row_lineage_prune_delete_test.go | 49 ++++++++++-----
table/row_lineage_rewrite_test.go | 2 +-
table/row_lineage_scan_test.go | 16 ++---
table/transaction.go | 107 +++++++++++++++++----------------
7 files changed, 158 insertions(+), 115 deletions(-)
diff --git a/table/data_file_meta_test.go b/table/data_file_meta_test.go
index b1101ea46..9ca145ed8 100644
--- a/table/data_file_meta_test.go
+++ b/table/data_file_meta_test.go
@@ -1228,7 +1228,7 @@ func
TestDataFileFromMetadata_V3RewriteAcceptsNilFirstRowID(t *testing.T) {
}
func TestDataFileFromMetadata_EndToEndRowDelta(t *testing.T) {
- tbl := newRowDeltaCommitTestTableVersion(t, 3)
+ tbl := newRowDeltaCommitTestTableVersion(t, 2)
sch := tbl.Schema()
const dataPath = "s3://bucket/test/data/insert.parquet"
diff --git a/table/row_delta.go b/table/row_delta.go
index bc37809f9..0e9508528 100644
--- a/table/row_delta.go
+++ b/table/row_delta.go
@@ -198,8 +198,13 @@ func (rd *RowDelta) Commit(ctx context.Context) error {
ct, f.FilePath())
}
- if err := validateDeletionVectorFormatVersion(f,
meta.formatVersion, "row delta"); err != nil {
- return err
+ if IsDeletionVector(f) {
+ if err := validateDeletionVectorToAdd(f,
meta.formatVersion, "row delta"); err != nil {
+ return err
+ }
+ } else if ct == iceberg.EntryContentPosDeletes &&
meta.formatVersion >= 3 {
+ return fmt.Errorf("position delete file %s must be a
deletion vector for v%d table for row delta",
+ f.FilePath(), meta.formatVersion)
}
// Equality delete files must declare which columns form the
delete key,
diff --git a/table/row_delta_test.go b/table/row_delta_test.go
index 38f791217..237f0457c 100644
--- a/table/row_delta_test.go
+++ b/table/row_delta_test.go
@@ -1108,66 +1108,86 @@ func TestRowDeltaRemoveDeletesFailsInsteadOfReplaying(t
*testing.T) {
"the data file must carry exactly one live DV: the peer's")
}
-func buildPuffinPosDeleteWithoutRef(t *testing.T, path string)
iceberg.DataFile {
- t.Helper()
-
- b, err := iceberg.NewDataFileBuilder(*iceberg.UnpartitionedSpec,
iceberg.EntryContentPosDeletes,
- path, iceberg.PuffinFile, nil, nil, nil, 2, 128)
- require.NoError(t, err)
-
- return b.Build()
-}
-
-func TestRowDeltaRejectsDeletionVectorBelowV3(t *testing.T) {
+func TestRowDeltaRejectsInvalidDeletionVector(t *testing.T) {
const (
dataPath = "s3://bucket/data/insert.parquet"
dvPath = "s3://bucket/data/dv-001.puffin"
)
+ offset, length := int64(4), int64(64)
+ negativeOffset, zeroLength := int64(-1), int64(0)
tests := []struct {
name string
formatVersion int
rows []iceberg.DataFile
- deletes func(*testing.T) []iceberg.DataFile
+ deletes []iceberg.DataFile
errContains string
}{
{
name: "deletion vector alone on v2",
formatVersion: 2,
- deletes: func(t *testing.T) []iceberg.DataFile {
- return []iceberg.DataFile{buildDVFile(t,
dvPath, dataPath)}
- },
- errContains: "requires table format version >= 3",
+ deletes: []iceberg.DataFile{buildDVFile(t,
dvPath, dataPath)},
+ errContains: "requires table format version >= 3",
},
{
name: "deletion vector beside valid files on
v2",
formatVersion: 2,
rows: []iceberg.DataFile{buildDataFile(t,
dataPath)},
- deletes: func(t *testing.T) []iceberg.DataFile {
- return []iceberg.DataFile{
- buildPosDeleteFile(t,
"s3://bucket/data/pos-del.parquet"),
- buildEqDeleteFile(t,
"s3://bucket/data/eq-del.parquet", []int{1}),
- buildDVFile(t, dvPath, dataPath),
- }
+ deletes: []iceberg.DataFile{
+ buildPosDeleteFile(t,
"s3://bucket/data/pos-del.parquet"),
+ buildEqDeleteFile(t,
"s3://bucket/data/eq-del.parquet", []int{1}),
+ buildDVFile(t, dvPath, dataPath),
},
errContains: "requires table format version >= 3",
},
{
- name: "deletion vector without referenced data
file on v2",
+ name: "format version check precedes field
checks on v2",
formatVersion: 2,
- deletes: func(t *testing.T) []iceberg.DataFile {
- return
[]iceberg.DataFile{buildPuffinPosDeleteWithoutRef(t, dvPath)}
- },
- errContains: "requires table format version >= 3",
+ deletes:
[]iceberg.DataFile{newRewriteDeletionVector(t, dvPath, "", nil, nil)},
+ errContains: "requires table format version >= 3",
},
{
// v1 rejects every delete file before the DV check is
reached.
name: "deletion vector on v1",
formatVersion: 1,
- deletes: func(t *testing.T) []iceberg.DataFile {
- return []iceberg.DataFile{buildDVFile(t,
dvPath, dataPath)}
- },
- errContains: "format version >= 2",
+ deletes: []iceberg.DataFile{buildDVFile(t,
dvPath, dataPath)},
+ errContains: "format version >= 2",
+ },
+ {
+ name: "missing referenced data file on v3",
+ formatVersion: 3,
+ deletes:
[]iceberg.DataFile{newRewriteDeletionVector(t, dvPath, "", &offset, &length)},
+ errContains: "missing referenced_data_file",
+ },
+ {
+ name: "missing content offset on v3",
+ formatVersion: 3,
+ deletes:
[]iceberg.DataFile{newRewriteDeletionVector(t, dvPath, dataPath, nil, &length)},
+ errContains: "missing content_offset",
+ },
+ {
+ name: "negative content offset on v3",
+ formatVersion: 3,
+ deletes:
[]iceberg.DataFile{newRewriteDeletionVector(t, dvPath, dataPath,
&negativeOffset, &length)},
+ errContains: "invalid content_offset -1",
+ },
+ {
+ name: "missing content size on v3",
+ formatVersion: 3,
+ deletes:
[]iceberg.DataFile{newRewriteDeletionVector(t, dvPath, dataPath, &offset, nil)},
+ errContains: "missing content_size_in_bytes",
+ },
+ {
+ name: "nonpositive content size on v3",
+ formatVersion: 3,
+ deletes:
[]iceberg.DataFile{newRewriteDeletionVector(t, dvPath, dataPath, &offset,
&zeroLength)},
+ errContains: "invalid content_size_in_bytes 0",
+ },
+ {
+ name: "parquet position delete on v3",
+ formatVersion: 3,
+ deletes: []iceberg.DataFile{buildPosDeleteFile(t,
"s3://bucket/data/pos-del.parquet")},
+ errContains: "must be a deletion vector for v3 table",
},
}
@@ -1175,7 +1195,7 @@ func TestRowDeltaRejectsDeletionVectorBelowV3(t
*testing.T) {
t.Run(tt.name, func(t *testing.T) {
tbl := newRowDeltaCommitTestTableVersion(t,
tt.formatVersion)
- rd :=
tbl.NewTransaction().NewRowDelta(nil).AddRows(tt.rows...).AddDeletes(tt.deletes(t)...)
+ rd :=
tbl.NewTransaction().NewRowDelta(nil).AddRows(tt.rows...).AddDeletes(tt.deletes...)
err := rd.Commit(t.Context())
require.Error(t, err)
@@ -1235,13 +1255,15 @@ func TestRowDeltaRemoveDeletesRequiresV3(t *testing.T) {
// DV) while adding a valid DV.
// Assertion: Commit fails identifying the live entry as not a DV.
func TestRowDeltaRemoveDeletesRejectsNonDV(t *testing.T) {
- tbl := newRowDeltaCommitTestTableVersion(t, 3)
+ tbl := newRowDeltaCommitTestTableVersion(t, 2)
location := tbl.Location()
+ // v3 rejects new Parquet pos-deletes, so the live one predates an
upgrade.
posDelPath := location + "/data/pos-del-001.parquet"
tx := tbl.NewTransaction()
require.NoError(t, tx.NewRowDelta(nil).
AddDeletes(buildPosDeleteFile(t,
posDelPath)).Commit(t.Context()))
+ require.NoError(t, tx.UpgradeFormatVersion(3))
tbl, err := tx.Commit(t.Context())
require.NoError(t, err)
diff --git a/table/row_lineage_prune_delete_test.go
b/table/row_lineage_prune_delete_test.go
index 9710a6b9f..1716b2fc1 100644
--- a/table/row_lineage_prune_delete_test.go
+++ b/table/row_lineage_prune_delete_test.go
@@ -43,12 +43,13 @@ import (
// test prunes a leading row group and deletes a row that lives only in the
// surviving group.
-// TestScanPruningWithPositionalDeletes covers a Parquet positional delete
whose
-// target (pos 6 = id=7) is in the second row group while the filter prunes the
-// first. The deleted row must vanish and survivors must keep their _row_id.
+// TestScanPruningWithPositionalDeletes covers a legacy Parquet positional
delete,
+// committed before the v3 upgrade, whose target (pos 6 = id=7) is in the
second row group
+// while the filter prunes the first.
+// The deleted row must vanish and survivors must keep their _row_id.
func TestScanPruningWithPositionalDeletes(t *testing.T) {
ctx := context.Background()
- tbl := buildTwoRowGroupV3Table(t)
+ tbl := buildTwoRowGroupTable(t, newV2RowLineageTestTable(t))
tasks, err := tbl.Scan().PlanFiles(ctx)
require.NoError(t, err)
@@ -64,10 +65,7 @@ func TestScanPruningWithPositionalDeletes(t *testing.T) {
*iceberg.UnpartitionedSpec, iceberg.EntryContentPosDeletes,
posDelPath, iceberg.ParquetFile, nil, nil, nil, 1, 256)
require.NoError(t, err)
- tx := tbl.NewTransaction()
- require.NoError(t,
tx.NewRowDelta(nil).AddDeletes(b.Build()).Commit(ctx))
- tbl, err = tx.Commit(ctx)
- require.NoError(t, err)
+ tbl = commitLegacyPosDelete(t, tbl, b.Build())
filter :=
table.WithRowFilter(iceberg.GreaterThan(iceberg.Reference("id"), int64(5)))
@@ -85,7 +83,7 @@ func TestScanPruningWithPositionalDeletes(t *testing.T) {
// drops pos 6 (id=7) in the surviving second row group while the first is
pruned.
func TestScanPruningWithDeletionVector(t *testing.T) {
ctx := context.Background()
- tbl := buildTwoRowGroupV3Table(t)
+ tbl := buildTwoRowGroupTable(t, newV3RowLineageTestTable(t))
tasks, err := tbl.Scan().PlanFiles(ctx)
require.NoError(t, err)
@@ -119,7 +117,7 @@ func TestScanPruningWithDeletionVector(t *testing.T) {
// must keep advancing across batch boundaries, not reset per batch.
func TestScanPruningSurvivingGroupSpansBatches(t *testing.T) {
ctx := context.Background()
- tbl := buildTwoRowGroupV3Table(t)
+ tbl := buildTwoRowGroupTable(t, newV3RowLineageTestTable(t))
tx := tbl.NewTransaction()
require.NoError(t,
tx.SetProperties(iceberg.Properties{table.ParquetBatchSizeKey: "2"}))
@@ -133,7 +131,7 @@ func TestScanPruningSurvivingGroupSpansBatches(t
*testing.T) {
func TestScanPruningAllRowGroupsCompletes(t *testing.T) {
ctx := context.Background()
- tbl := buildTwoRowGroupV3Table(t)
+ tbl := buildTwoRowGroupTable(t, newV3RowLineageTestTable(t))
_, itr, err := tbl.Scan(
table.WithRowLineage(),
@@ -222,12 +220,11 @@ func
TestReadTaskDeletionVectorSupersedesPositionalDeletes(t *testing.T) {
// scan cannot prove bloom skipped a group, since a correct scan emits the same
// rows whether or not the empty group was physically read.
-// buildTwoRowGroupV3Table writes a v3 row-lineage table with ids 1..10 across
two
-// 5-row row groups, so an id>5 filter prunes the first group via stats.
-func buildTwoRowGroupV3Table(t *testing.T) *table.Table {
+// buildTwoRowGroupTable appends ids 1..10 to tbl across two 5-row row groups,
+// so an id>5 filter prunes the first group via stats.
+func buildTwoRowGroupTable(t *testing.T, tbl *table.Table) *table.Table {
t.Helper()
ctx := context.Background()
- tbl := newV3RowLineageTestTable(t)
tx := tbl.NewTransaction()
require.NoError(t,
tx.SetProperties(iceberg.Properties{table.ParquetRowGroupLimitKey: "5"}))
@@ -255,6 +252,28 @@ func buildTwoRowGroupV3Table(t *testing.T) *table.Table {
return tbl
}
+// commitLegacyPosDelete adds posDel while tbl is still v2, then upgrades it
to v3.
+// The no-op Delete must keep committing a v3 snapshot: that snapshot gives
the existing rows a _row_id.
+func commitLegacyPosDelete(t *testing.T, tbl *table.Table, posDel
iceberg.DataFile) *table.Table {
+ t.Helper()
+ ctx := context.Background()
+
+ tx := tbl.NewTransaction()
+ require.NoError(t, tx.NewRowDelta(nil).AddDeletes(posDel).Commit(ctx))
+ require.NoError(t, tx.UpgradeFormatVersion(3))
+ require.NoError(t, tx.Delete(ctx, iceberg.AlwaysFalse{}, nil))
+ tbl, err := tx.Commit(ctx)
+ require.NoError(t, err)
+
+ tasks, err := tbl.Scan().PlanFiles(ctx)
+ require.NoError(t, err)
+ for _, task := range tasks {
+ require.NotNil(t, task.FirstRowID, "upgraded data file %s must
inherit a first_row_id", task.File.FilePath())
+ }
+
+ return tbl
+}
+
// idsMatchingFilter scans without row lineage and returns the surviving ids,
sorted.
func idsMatchingFilter(t *testing.T, tbl *table.Table, opts
...table.ScanOption) []int64 {
t.Helper()
diff --git a/table/row_lineage_rewrite_test.go
b/table/row_lineage_rewrite_test.go
index 2c9b87c7b..d6f7e5ae2 100644
--- a/table/row_lineage_rewrite_test.go
+++ b/table/row_lineage_rewrite_test.go
@@ -302,7 +302,7 @@ func TestCoWRewriteDoesNotPruneMissingInitialDefault(t
*testing.T) {
func TestCoWRewriteNormalizesTheSurvivorComplement(t *testing.T) {
ctx := context.Background()
- tbl := buildTwoRowGroupV3Table(t)
+ tbl := buildTwoRowGroupTable(t, newV3RowLineageTestTable(t))
var err error
tbl, err = tbl.Delete(ctx,
diff --git a/table/row_lineage_scan_test.go b/table/row_lineage_scan_test.go
index 909dfdd79..619c872e1 100644
--- a/table/row_lineage_scan_test.go
+++ b/table/row_lineage_scan_test.go
@@ -120,11 +120,12 @@ func TestScanRowLineagePreservedThroughEqualityDeletes(t
*testing.T) {
"equality-deleted rows must not renumber surviving rows'
_row_id")
}
-// TestScanRowLineagePreservedThroughPositionalDeletes: a Parquet
position-delete
-// file drops rows during the scan; survivors must keep their original _row_id.
+// TestScanRowLineagePreservedThroughPositionalDeletes: a legacy Parquet
position-delete
+// file, committed before the v3 upgrade, drops rows during the scan;
survivors must
+// keep their original _row_id.
func TestScanRowLineagePreservedThroughPositionalDeletes(t *testing.T) {
ctx := context.Background()
- tbl := newV3RowLineageTestTable(t)
+ tbl := newV2RowLineageTestTable(t)
arrowSchema := arrow.NewSchema([]arrow.Field{
{Name: "id", Type: arrow.PrimitiveTypes.Int64, Nullable: false},
@@ -138,7 +139,6 @@ func TestScanRowLineagePreservedThroughPositionalDeletes(t
*testing.T) {
tbl, err = tbl.Append(ctx, array.NewTableReader(data, -1), nil)
require.NoError(t, err)
- require.Equal(t, map[int64]int64{1: 0, 2: 1, 3: 2, 4: 3, 5: 4},
rowIDByID(t, tbl))
tasks, err := tbl.Scan().PlanFiles(ctx)
require.NoError(t, err)
@@ -156,13 +156,7 @@ func TestScanRowLineagePreservedThroughPositionalDeletes(t
*testing.T) {
*iceberg.UnpartitionedSpec, iceberg.EntryContentPosDeletes,
posDelPath, iceberg.ParquetFile, nil, nil, nil, 2, 256)
require.NoError(t, err)
-
- tx := tbl.NewTransaction()
- rd := tx.NewRowDelta(nil)
- rd.AddDeletes(b.Build())
- require.NoError(t, rd.Commit(ctx))
- tbl, err = tx.Commit(ctx)
- require.NoError(t, err)
+ tbl = commitLegacyPosDelete(t, tbl, b.Build())
assert.Equal(t, map[int64]int64{1: 0, 3: 2, 5: 4}, rowIDByID(t, tbl),
"position-deleted rows must not renumber surviving rows'
_row_id")
diff --git a/table/transaction.go b/table/transaction.go
index 1b5508f24..6a96b7d5f 100644
--- a/table/transaction.go
+++ b/table/transaction.go
@@ -1189,9 +1189,31 @@ type deleteFilesToAddSet struct {
dvsByRef map[string]rewriteDeleteFileAddition
}
-func validateDeletionVectorFormatVersion(df iceberg.DataFile, formatVersion
int, operation string) error {
- if IsDeletionVector(df) && formatVersion < 3 {
- return fmt.Errorf("deletion vector %s requires table format
version >= 3 for %s", df.FilePath(), operation)
+// validateDeletionVectorToAdd is safe on any DataFile and no-ops on non-DVs.
+func validateDeletionVectorToAdd(df iceberg.DataFile, formatVersion int,
operation string) error {
+ if !IsDeletionVector(df) {
+ return nil
+ }
+ path := df.FilePath()
+ if formatVersion < 3 {
+ return fmt.Errorf("deletion vector %s requires table format
version >= 3 for %s", path, operation)
+ }
+ if ref := df.ReferencedDataFile(); ref == nil || *ref == "" {
+ return fmt.Errorf("deletion vector %s is missing
referenced_data_file for %s", path, operation)
+ }
+ offset := df.ContentOffset()
+ if offset == nil {
+ return fmt.Errorf("deletion vector %s is missing content_offset
for %s", path, operation)
+ }
+ if *offset < 0 {
+ return fmt.Errorf("deletion vector %s has invalid
content_offset %d for %s", path, *offset, operation)
+ }
+ length := df.ContentSizeInBytes()
+ if length == nil {
+ return fmt.Errorf("deletion vector %s is missing
content_size_in_bytes for %s", path, operation)
+ }
+ if *length <= 0 {
+ return fmt.Errorf("deletion vector %s has invalid
content_size_in_bytes %d for %s", path, *length, operation)
}
return nil
@@ -1265,10 +1287,6 @@ func (t *Transaction)
validateDeleteFilesToAdd(deleteFiles []rewriteDeleteFileAd
}
}
- if err := validateDeletionVectorFormatVersion(df,
meta.formatVersion, operation); err != nil {
- return nil, err
- }
-
if !IsDeletionVector(df) {
if meta.formatVersion >= 3 && df.ContentType() ==
iceberg.EntryContentPosDeletes {
return nil, fmt.Errorf("position delete file %s
must be a deletion vector for v%d table for %s",
@@ -1286,54 +1304,39 @@ func (t *Transaction)
validateDeleteFilesToAdd(deleteFiles []rewriteDeleteFileAd
continue
}
- if IsDeletionVector(df) {
- ref := df.ReferencedDataFile()
- if ref == nil || *ref == "" {
- return nil, fmt.Errorf("deletion vector to add
is missing referenced_data_file for %s", operation)
- }
- offset := df.ContentOffset()
- if offset == nil {
- return nil, fmt.Errorf("deletion vector %s is
missing content_offset for %s", path, operation)
- }
- if *offset < 0 {
- return nil, fmt.Errorf("deletion vector %s has
invalid content_offset %d for %s", path, *offset, operation)
- }
- length := df.ContentSizeInBytes()
- if length == nil {
- return nil, fmt.Errorf("deletion vector %s is
missing content_size_in_bytes for %s", path, operation)
- }
- if *length <= 0 {
- return nil, fmt.Errorf("deletion vector %s has
invalid content_size_in_bytes %d for %s", path, *length, operation)
- }
+ if err := validateDeletionVectorToAdd(df, meta.formatVersion,
operation); err != nil {
+ return nil, err
+ }
+ // non-nil: validateDeletionVectorToAdd checked all three.
+ ref, offset, length := df.ReferencedDataFile(),
df.ContentOffset(), df.ContentSizeInBytes()
- blob := deletionVectorBlobKey{path: path, offset:
*offset, length: *length}
- if setToAdd.dvBlobs == nil {
- setToAdd.dvBlobs =
make(map[deletionVectorBlobKey]struct{}, len(deleteFiles))
- }
- if _, ok := setToAdd.dvBlobs[blob]; ok {
- return nil, fmt.Errorf("deletion vector blob
identity must be unique for %s: %s at offset %d with length %d",
- operation, path, *offset, *length)
- }
- if _, ok := setToAdd.dvsByRef[*ref]; ok {
- return nil, fmt.Errorf("deletion vectors to add
must reference distinct data files for %s: %s",
- operation, *ref)
- }
- if pathAlreadyAdded {
- if _, ok := setToAdd.dvPaths[path]; !ok {
- return nil, fmt.Errorf("delete file
path %s cannot identify both a deletion vector container and a regular delete
file for %s",
- path, operation)
- }
- }
- if setToAdd.dvPaths == nil {
- setToAdd.dvPaths = make(map[string]struct{},
len(deleteFiles))
- }
- if setToAdd.dvsByRef == nil {
- setToAdd.dvsByRef =
make(map[string]rewriteDeleteFileAddition, len(deleteFiles))
+ blob := deletionVectorBlobKey{path: path, offset: *offset,
length: *length}
+ if setToAdd.dvBlobs == nil {
+ setToAdd.dvBlobs =
make(map[deletionVectorBlobKey]struct{}, len(deleteFiles))
+ }
+ if _, ok := setToAdd.dvBlobs[blob]; ok {
+ return nil, fmt.Errorf("deletion vector blob identity
must be unique for %s: %s at offset %d with length %d",
+ operation, path, *offset, *length)
+ }
+ if _, ok := setToAdd.dvsByRef[*ref]; ok {
+ return nil, fmt.Errorf("deletion vectors to add must
reference distinct data files for %s: %s",
+ operation, *ref)
+ }
+ if pathAlreadyAdded {
+ if _, ok := setToAdd.dvPaths[path]; !ok {
+ return nil, fmt.Errorf("delete file path %s
cannot identify both a deletion vector container and a regular delete file for
%s",
+ path, operation)
}
- setToAdd.dvPaths[path] = struct{}{}
- setToAdd.dvBlobs[blob] = struct{}{}
- setToAdd.dvsByRef[*ref] = addition
}
+ if setToAdd.dvPaths == nil {
+ setToAdd.dvPaths = make(map[string]struct{},
len(deleteFiles))
+ }
+ if setToAdd.dvsByRef == nil {
+ setToAdd.dvsByRef =
make(map[string]rewriteDeleteFileAddition, len(deleteFiles))
+ }
+ setToAdd.dvPaths[path] = struct{}{}
+ setToAdd.dvBlobs[blob] = struct{}{}
+ setToAdd.dvsByRef[*ref] = addition
}
return setToAdd, nil
