Compare commits

..

16 Commits

Author SHA1 Message Date
Andrew Kane
4f4286f74e Updated handling of range error to be consistent with real 2024-03-30 08:59:23 -07:00
Andrew Kane
11ea3d8483 Updated SQL comments [skip ci] 2024-03-30 08:29:01 -07:00
Andrew Kane
2c48e3edc2 Mark type-specific code 2024-03-29 14:01:48 -07:00
Andrew Kane
7d63bb4b98 Fixed flaky test [skip ci] 2024-03-29 11:00:16 -07:00
Andrew Kane
de410a2915 Use variable for max dimemsions [skip ci] 2024-03-29 10:57:16 -07:00
Andrew Kane
64aa99aa31 Added todo [skip ci] 2024-03-29 10:56:24 -07:00
Andrew Kane
997fa167da Removed vector-specific code from HNSW 2024-03-29 10:50:06 -07:00
Andrew Kane
67eec4edbf Improved tuning section [skip ci] 2024-03-27 22:04:28 -07:00
Andrew Kane
396090d8e0 Improved code [skip ci] 2024-03-27 21:38:22 -07:00
Andrew Kane
ba18942fcf Removed normvec from IVFFlat for simplicity (no difference in performance) 2024-03-27 16:41:17 -07:00
Andrew Kane
8e59455c3c Removed normvec for simplicity (no difference in performance) 2024-03-27 16:33:11 -07:00
Andrew Kane
bd50e3067d Updated readme [skip ci] 2024-03-27 14:14:49 -07:00
Andrew Kane
af9d4ad659 Updated readme [skip ci] 2024-03-27 14:12:08 -07:00
Andrew Kane
08abb63cbe Added notes about NULL vectors [skip ci] 2024-03-27 11:50:37 -07:00
Andrew Kane
06b8556a49 Revert "Updated readme [skip ci]"
This reverts commit 3f674c9994.
2024-03-25 23:33:46 -07:00
Andrew Kane
3f674c9994 Updated readme [skip ci] 2024-03-25 23:33:17 -07:00
22 changed files with 115 additions and 163 deletions

View File

@@ -1,7 +1,3 @@
## 0.7.0 (unreleased)
- Added `subvector` function
## 0.6.2 (2024-03-18)
- Reduced lock contention with parallel HNSW index builds

View File

@@ -415,7 +415,19 @@ You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python
### Tuning
Use a tool like [PgTune](https://pgtune.leopard.in.ua/) to set initial values for Postgres server parameters.
Use a tool like [PgTune](https://pgtune.leopard.in.ua/) to set initial values for Postgres server parameters. For instance, `shared_buffers` should typically be 25% of the servers memory. You can find the config file with:
```sql
SHOW config_file;
```
And check individual settings with:
```sql
SHOW shared_buffers;
```
Be sure to restart Postgres for changes to take effect.
### Loading
@@ -671,6 +683,8 @@ ALTER TABLE items ALTER COLUMN embedding SET STORAGE PLAIN;
Results are limited by the size of the dynamic candidate list (`hnsw.ef_search`). There may be even less results due to dead tuples or filtering conditions in the query. We recommend setting `hnsw.ef_search` to at least twice the `LIMIT` of the query. If you need more than 500 results, use an IVFFlat index instead.
Also, note that `NULL` vectors are not indexed (as well as zero vectors for cosine distance).
#### Why are there less results for a query after adding an IVFFlat index?
The index was likely created with too little data for the number of lists. Drop the index until the table has more data.
@@ -681,11 +695,13 @@ DROP INDEX index_name;
Results can also be limited by the number of probes (`ivfflat.probes`).
Also, note that `NULL` vectors are not indexed (as well as zero vectors for cosine distance).
## Reference
### Vector Type
Each vector takes `4 * dimensions + 8` bytes of storage. Each element is a single precision floating-point number (like the `real` type in Postgres), and all elements must be finite (no `NaN`, `Infinity` or `-Infinity`). Vectors can have up to 16,000 dimensions.
Each vector takes `4 * dimensions + 8` bytes of storage. Each element is a single-precision floating-point number (like the `real` type in Postgres), and all elements must be finite (no `NaN`, `Infinity` or `-Infinity`). Vectors can have up to 16,000 dimensions.
### Vector Operators
@@ -706,11 +722,10 @@ cosine_distance(vector, vector) → double precision | cosine distance |
inner_product(vector, vector) → double precision | inner product |
l2_distance(vector, vector) → double precision | Euclidean distance |
l1_distance(vector, vector) → double precision | taxicab distance | 0.5.0
subvector(vector, integer, integer) → vector | subvector | 0.7.0 [unreleased]
vector_dims(vector) → integer | number of dimensions |
vector_norm(vector) → double precision | Euclidean norm |
### Aggregate Functions
### Vector Aggregate Functions
Function | Description | Added
--- | --- | ---

View File

@@ -1,5 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.0'" to load this file. \quit
CREATE FUNCTION subvector(vector, int, int) RETURNS vector
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;

View File

@@ -1,7 +1,7 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "CREATE EXTENSION vector" to load this file. \quit
-- type
-- vector type
CREATE TYPE vector;
@@ -29,7 +29,7 @@ CREATE TYPE vector (
STORAGE = external
);
-- functions
-- vector functions
CREATE FUNCTION l2_distance(vector, vector) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
@@ -58,10 +58,7 @@ CREATE FUNCTION vector_sub(vector, vector) RETURNS vector
CREATE FUNCTION vector_mul(vector, vector) RETURNS vector
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION subvector(vector, int, int) RETURNS vector
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- private functions
-- vector private functions
CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
@@ -102,7 +99,7 @@ CREATE FUNCTION vector_avg(double precision[]) RETURNS vector
CREATE FUNCTION vector_combine(double precision[], double precision[]) RETURNS double precision[]
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- aggregates
-- vector aggregates
CREATE AGGREGATE avg(vector) (
SFUNC = vector_accum,
@@ -120,7 +117,7 @@ CREATE AGGREGATE sum(vector) (
PARALLEL = SAFE
);
-- cast functions
-- vector cast functions
CREATE FUNCTION vector(vector, integer, boolean) RETURNS vector
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
@@ -140,7 +137,7 @@ CREATE FUNCTION array_to_vector(numeric[], integer, boolean) RETURNS vector
CREATE FUNCTION vector_to_float4(vector, integer, boolean) RETURNS real[]
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- casts
-- vector casts
CREATE CAST (vector AS vector)
WITH FUNCTION vector(vector, integer, boolean) AS IMPLICIT;
@@ -160,7 +157,7 @@ CREATE CAST (double precision[] AS vector)
CREATE CAST (numeric[] AS vector)
WITH FUNCTION array_to_vector(numeric[], integer, boolean) AS ASSIGNMENT;
-- operators
-- vector operators
CREATE OPERATOR <-> (
LEFTARG = vector, RIGHTARG = vector, PROCEDURE = l2_distance,
@@ -243,7 +240,7 @@ CREATE ACCESS METHOD hnsw TYPE INDEX HANDLER hnswhandler;
COMMENT ON ACCESS METHOD hnsw IS 'hnsw index access method';
-- opclasses
-- vector opclasses
CREATE OPERATOR CLASS vector_ops
DEFAULT FOR TYPE vector USING btree AS

View File

@@ -55,6 +55,11 @@
#define HNSW_UPDATE_ENTRY_GREATER 1
#define HNSW_UPDATE_ENTRY_ALWAYS 2
typedef enum HnswType
{
HNSW_TYPE_VECTOR
} HnswType;
/* Build phases */
/* PROGRESS_CREATEIDX_SUBPHASE_INITIALIZE is 1 */
#define PROGRESS_HNSW_PHASE_LOAD 2
@@ -242,6 +247,7 @@ typedef struct HnswBuildState
Relation index;
IndexInfo *indexInfo;
ForkNumber forkNum;
HnswType type;
/* Settings */
int dimensions;
@@ -262,7 +268,6 @@ typedef struct HnswBuildState
HnswGraph *graph;
double ml;
int maxLevel;
Vector *normvec;
/* Memory */
MemoryContext graphCtx;
@@ -367,7 +372,8 @@ typedef struct HnswVacuumState
int HnswGetM(Relation index);
int HnswGetEfConstruction(Relation index);
FmgrInfo *HnswOptionalProcInfo(Relation index, uint16 procnum);
bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
HnswType HnswGetType(Relation index);
bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, HnswType type);
Buffer HnswNewBuffer(Relation index, ForkNumber forkNum);
void HnswInitPage(Buffer buf, Page page);
void HnswInit(void);

View File

@@ -489,7 +489,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
/* Normalize if needed */
if (buildstate->normprocinfo != NULL)
{
if (!HnswNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec))
if (!HnswNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->type))
return false;
}
@@ -671,10 +671,13 @@ HnswSharedMemoryAlloc(Size size, void *state)
static void
InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, IndexInfo *indexInfo, ForkNumber forkNum)
{
int maxDimensions = HNSW_MAX_DIM;
buildstate->heap = heap;
buildstate->index = index;
buildstate->indexInfo = indexInfo;
buildstate->forkNum = forkNum;
buildstate->type = HnswGetType(index);
buildstate->m = HnswGetM(index);
buildstate->efConstruction = HnswGetEfConstruction(index);
@@ -684,8 +687,8 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
if (buildstate->dimensions < 0)
elog(ERROR, "column does not have dimensions");
if (buildstate->dimensions > HNSW_MAX_DIM)
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", HNSW_MAX_DIM);
if (buildstate->dimensions > maxDimensions)
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", maxDimensions);
if (buildstate->efConstruction < 2 * buildstate->m)
elog(ERROR, "ef_construction must be greater than or equal to 2 * m");
@@ -703,9 +706,6 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
buildstate->ml = HnswGetMl(buildstate->m);
buildstate->maxLevel = HnswGetMaxLevel(buildstate->m);
/* Reuse for each tuple */
buildstate->normvec = InitVector(buildstate->dimensions);
buildstate->graphCtx = GenerationContextCreate(CurrentMemoryContext,
"Hnsw build graph context",
#if PG_VERSION_NUM >= 150000
@@ -729,7 +729,6 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
static void
FreeBuildState(HnswBuildState * buildstate)
{
pfree(buildstate->normvec);
MemoryContextDelete(buildstate->graphCtx);
MemoryContextDelete(buildstate->tmpCtx);
}

View File

@@ -622,7 +622,7 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
if (normprocinfo != NULL)
{
if (!HnswNormValue(normprocinfo, collation, &value, NULL))
if (!HnswNormValue(normprocinfo, collation, &value, HnswGetType(index)))
return;
}

View File

@@ -40,29 +40,6 @@ GetScanItems(IndexScanDesc scan, Datum q)
return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
}
/*
* Get dimensions from metapage
*/
static int
GetDimensions(Relation index)
{
Buffer buf;
Page page;
HnswMetaPage metap;
int dimensions;
buf = ReadBuffer(index, HNSW_METAPAGE_BLKNO);
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
metap = HnswPageGetMeta(page);
dimensions = metap->dimensions;
UnlockReleaseBuffer(buf);
return dimensions;
}
/*
* Get scan value
*/
@@ -73,7 +50,7 @@ GetScanValue(IndexScanDesc scan)
Datum value;
if (scan->orderByData->sk_flags & SK_ISNULL)
value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
value = PointerGetDatum(NULL);
else
{
value = scan->orderByData->sk_argument;
@@ -84,7 +61,7 @@ GetScanValue(IndexScanDesc scan)
/* Fine if normalization fails */
if (so->normprocinfo != NULL)
HnswNormValue(so->normprocinfo, so->collation, &value, NULL);
HnswNormValue(so->normprocinfo, so->collation, &value, HnswGetType(scan->indexRelation));
}
return value;

View File

@@ -149,6 +149,15 @@ HnswOptionalProcInfo(Relation index, uint16 procnum)
return index_getprocinfo(index, 1, procnum);
}
/*
* Get vector type
*/
HnswType
HnswGetType(Relation index)
{
return HNSW_TYPE_VECTOR;
}
/*
* Divide by the norm
*
@@ -158,21 +167,25 @@ HnswOptionalProcInfo(Relation index, uint16 procnum)
* if it's different than the original value
*/
bool
HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result)
HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, HnswType type)
{
double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
if (norm > 0)
{
Vector *v = DatumGetVector(*value);
/* TODO Remove vector-specific code */
if (type == HNSW_TYPE_VECTOR)
{
Vector *v = DatumGetVector(*value);
Vector *result = InitVector(v->dim);
if (result == NULL)
result = InitVector(v->dim);
for (int i = 0; i < v->dim; i++)
result->x[i] = v->x[i] / norm;
for (int i = 0; i < v->dim; i++)
result->x[i] = v->x[i] / norm;
*value = PointerGetDatum(result);
*value = PointerGetDatum(result);
}
else
elog(ERROR, "Unsupported type");
return true;
}
@@ -575,7 +588,12 @@ HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index,
/* Calculate distance */
if (distance != NULL)
*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->data)));
{
if (DatumGetPointer(*q) == NULL)
*distance = 0;
else
*distance = (float) DatumGetFloat8(FunctionCall2Coll(procinfo, collation, *q, PointerGetDatum(&etup->data)));
}
UnlockReleaseBuffer(buf);
}

View File

@@ -57,7 +57,7 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
*/
if (buildstate->kmeansnormprocinfo != NULL)
{
if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value, buildstate->normvec))
if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value))
return;
}
@@ -105,7 +105,7 @@ SampleCallback(Relation index, CALLBACK_ITEM_POINTER, Datum *values,
oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx);
/* Add sample */
AddSample(values, state);
AddSample(values, buildstate);
/* Reset memory context */
MemoryContextSwitchTo(oldCtx);
@@ -153,7 +153,7 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
/* Normalize if needed */
if (buildstate->normprocinfo != NULL)
{
if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec))
if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value))
return;
}
@@ -356,9 +356,6 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions);
buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
/* Reuse for each tuple */
buildstate->normvec = InitVector(buildstate->dimensions);
buildstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Ivfflat build temporary context",
ALLOCSET_DEFAULT_SIZES);
@@ -380,7 +377,6 @@ FreeBuildState(IvfflatBuildState * buildstate)
{
VectorArrayFree(buildstate->centers);
pfree(buildstate->listInfo);
pfree(buildstate->normvec);
#ifdef IVFFLAT_KMEANS_DEBUG
pfree(buildstate->listSums);

View File

@@ -172,7 +172,6 @@ typedef struct IvfflatBuildState
VectorArray samples;
VectorArray centers;
ListInfo *listInfo;
Vector *normvec;
#ifdef IVFFLAT_KMEANS_DEBUG
double inertia;
@@ -267,7 +266,7 @@ void VectorArrayFree(VectorArray arr);
void PrintVectorArray(char *msg, VectorArray arr);
void IvfflatKmeans(Relation index, VectorArray samples, VectorArray centers);
FmgrInfo *IvfflatOptionalProcInfo(Relation index, uint16 procnum);
bool IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
bool IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value);
int IvfflatGetLists(Relation index);
void IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions);
void IvfflatUpdateList(Relation index, ListInfo listInfo, BlockNumber insertPage, BlockNumber originalInsertPage, BlockNumber startPage, ForkNumber forkNum);

View File

@@ -85,7 +85,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
if (normprocinfo != NULL)
{
if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value, NULL))
if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value))
return;
}

View File

@@ -293,7 +293,7 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
/* Fine if normalization fails */
if (so->normprocinfo != NULL)
IvfflatNormValue(so->normprocinfo, so->collation, &value, NULL);
IvfflatNormValue(so->normprocinfo, so->collation, &value);
}
IvfflatBench("GetScanLists", GetScanLists(scan, value));

View File

@@ -75,16 +75,14 @@ IvfflatOptionalProcInfo(Relation index, uint16 procnum)
* if it's different than the original value
*/
bool
IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result)
IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value)
{
double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
if (norm > 0)
{
Vector *v = DatumGetVector(*value);
if (result == NULL)
result = InitVector(v->dim);
Vector *result = InitVector(v->dim);
for (int i = 0; i < v->dim; i++)
result->x[i] = v->x[i] / norm;

View File

@@ -197,6 +197,8 @@ vector_in(PG_FUNCTION_ARGS)
while (pt != NULL && *stringEnd != ']')
{
float val;
if (dim == VECTOR_MAX_DIM)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
@@ -212,15 +214,23 @@ vector_in(PG_FUNCTION_ARGS)
errmsg("invalid input syntax for type vector: \"%s\"", lit)));
/* Use strtof like float4in to avoid a double-rounding problem */
x[dim] = strtof(pt, &stringEnd);
CheckElement(x[dim]);
dim++;
errno = 0;
val = strtof(pt, &stringEnd);
if (stringEnd == pt)
ereport(ERROR,
(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
errmsg("invalid input syntax for type vector: \"%s\"", lit)));
/* Check for range error like float4in */
if (errno == ERANGE && (val == 0 || isinf(val)))
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("\"%s\" is out of range for type vector", pt)));
CheckElement(val);
x[dim++] = val;
while (vector_isspace(*stringEnd))
stringEnd++;
@@ -860,37 +870,6 @@ vector_mul(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
/*
* Get a subset of a vector
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(subvector);
Datum
subvector(PG_FUNCTION_ARGS)
{
Vector *a = PG_GETARG_VECTOR_P(0);
int32 start = PG_GETARG_INT32(1);
int32 count = PG_GETARG_INT32(2);
int32 end = start + count;
float *ax = a->x;
Vector *result;
int dim;
if (start < 1)
start = 1;
if (end > a->dim)
end = a->dim + 1;
dim = end - start;
CheckDim(dim);
result = InitVector(dim);
for (int i = 0; i < dim; i++)
result->x[i] = ax[start - 1 + i];
PG_RETURN_POINTER(result);
}
/*
* Internal helper to compare vectors
*/

View File

@@ -208,34 +208,6 @@ SELECT l1_distance('[3e38]', '[-3e38]');
Infinity
(1 row)
SELECT subvector('[1,2,3,4,5]', 1, 3);
subvector
-----------
[1,2,3]
(1 row)
SELECT subvector('[1,2,3,4,5]', 3, 2);
subvector
-----------
[3,4]
(1 row)
SELECT subvector('[1,2,3,4,5]', -1, 3);
subvector
-----------
[1]
(1 row)
SELECT subvector('[1,2,3,4,5]', 3, 9);
subvector
-----------
[3,4,5]
(1 row)
SELECT subvector('[1,2,3,4,5]', 1, 0);
ERROR: vector must have at least 1 dimension
SELECT subvector('[1,2,3,4,5]', -1, 2);
ERROR: vector must have at least 1 dimension
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
avg
-----------

View File

@@ -12,14 +12,11 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
[0,0,0]
(4 rows)
SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
val
---------
[0,0,0]
[1,1,1]
[1,2,3]
[1,2,4]
(4 rows)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector)) t2;
count
-------
4
(1 row)
SELECT COUNT(*) FROM t;
count

View File

@@ -63,9 +63,21 @@ SELECT '[1.5e-38,-1.5e-38]'::vector;
(1 row)
SELECT '[4e38,1]'::vector;
ERROR: infinite value not allowed in vector
ERROR: "4e38" is out of range for type vector
LINE 1: SELECT '[4e38,1]'::vector;
^
SELECT '[-4e38,1]'::vector;
ERROR: "-4e38" is out of range for type vector
LINE 1: SELECT '[-4e38,1]'::vector;
^
SELECT '[1e-46,1]'::vector;
ERROR: "1e-46" is out of range for type vector
LINE 1: SELECT '[1e-46,1]'::vector;
^
SELECT '[-1e-46,1]'::vector;
ERROR: "-1e-46" is out of range for type vector
LINE 1: SELECT '[-1e-46,1]'::vector;
^
SELECT '[1,2,3'::vector;
ERROR: malformed vector literal: "[1,2,3"
LINE 1: SELECT '[1,2,3'::vector;

View File

@@ -48,13 +48,6 @@ SELECT l1_distance('[0,0]', '[0,1]');
SELECT l1_distance('[1,2]', '[3]');
SELECT l1_distance('[3e38]', '[-3e38]');
SELECT subvector('[1,2,3,4,5]', 1, 3);
SELECT subvector('[1,2,3,4,5]', 3, 2);
SELECT subvector('[1,2,3,4,5]', -1, 3);
SELECT subvector('[1,2,3,4,5]', 3, 9);
SELECT subvector('[1,2,3,4,5]', 1, 0);
SELECT subvector('[1,2,3,4,5]', -1, 2);
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;

View File

@@ -7,7 +7,7 @@ CREATE INDEX ON t USING hnsw (val vector_l2_ops);
INSERT INTO t (val) VALUES ('[1,2,4]');
SELECT * FROM t ORDER BY val <-> '[3,3,3]';
SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector);
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <-> (SELECT NULL::vector)) t2;
SELECT COUNT(*) FROM t;
TRUNCATE t;

View File

@@ -11,6 +11,9 @@ SELECT '[1.5e38,-1.5e38]'::vector;
SELECT '[1.5e+38,-1.5e+38]'::vector;
SELECT '[1.5e-38,-1.5e-38]'::vector;
SELECT '[4e38,1]'::vector;
SELECT '[-4e38,1]'::vector;
SELECT '[1e-46,1]'::vector;
SELECT '[-1e-46,1]'::vector;
SELECT '[1,2,3'::vector;
SELECT '[1,2,3]9'::vector;
SELECT '1,2,3'::vector;

View File

@@ -86,7 +86,7 @@ foreach (@queries)
push(@expected, $res);
}
test_recall(0.20, $limit, "before vacuum");
test_recall(0.19, $limit, "before vacuum");
test_recall(0.95, 100, "before vacuum");
# TODO Test concurrent inserts with vacuum