/*
* Finish the open by positioning the next read and saving information.
*
* file - The open file.
* filePathName - name of the segment file to open.
* version - AO table format version the file is in.
* logicalEof - snapshot version of the EOF value to use as the read end
* of the segment file.
*/
static void
AppendOnlyStorageRead_FinishOpenFile(AppendOnlyStorageRead *storageRead,
File file,
char *filePathName,
int version,
int64 logicalEof)
{
int64 seekResult;
MemoryContext oldMemoryContext;
int segmentFileNameLen;
AORelationVersion_CheckValid(version);
/*
* Seek to the beginning of the file.
*/
seekResult = FileSeek(file, 0, SEEK_SET);
if (seekResult != 0)
{
FileClose(file);
ereport(ERROR,
(errcode(ERRCODE_IO_ERROR),
errmsg("Append-only Storage Read error on segment file '%s' for relation '%s'. FileSeek offset = 0. Error code = %d (%s)",
filePathName,
storageRead->relationName,
(int) seekResult,
strerror((int) seekResult))));
}
storageRead->file = file;
storageRead->formatVersion = version;
/*
* When reading multiple segment files, we throw away the old segment file
* name strings.
*/
oldMemoryContext = MemoryContextSwitchTo(storageRead->memoryContext);
if (storageRead->segmentFileName != NULL)
pfree(storageRead->segmentFileName);
segmentFileNameLen = strlen(filePathName);
storageRead->segmentFileName = (char *) palloc(segmentFileNameLen + 1);
memcpy(storageRead->segmentFileName, filePathName, segmentFileNameLen + 1);
/* Allocation is done. Go back to caller memory-context. */
MemoryContextSwitchTo(oldMemoryContext);
storageRead->logicalEof = logicalEof;
BufferedReadSetFile(
&storageRead->bufferedRead,
storageRead->file,
storageRead->segmentFileName,
logicalEof);
}
/**
*begin scanning of a parquet relation
*/
ParquetScanDesc
parquet_beginscan(
Relation relation,
Snapshot parquetMetaDataSnapshot,
TupleDesc relationTupleDesc,
bool *proj)
{
ParquetScanDesc scan;
AppendOnlyEntry *aoEntry;
AppendOnlyStorageAttributes *attr;
/*
* increment relation ref count while scanning relation
*
* This is just to make really sure the relcache entry won't go away while
* the scan has a pointer to it. Caller should be holding the rel open
* anyway, so this is redundant in all normal scenarios...
*/
RelationIncrementReferenceCount(relation);
/* allocate scan descriptor */
scan = (ParquetScanDescData *)palloc0(sizeof(ParquetScanDescData));
/*
* Get the pg_appendonly information for this table
*/
aoEntry = GetAppendOnlyEntry(RelationGetRelid(relation), parquetMetaDataSnapshot);
scan->aoEntry = aoEntry;
Assert(aoEntry->majorversion == 1 && aoEntry->minorversion == 0);
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FailQeWhenBeginParquetScan,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
/*
* initialize the scan descriptor
*/
scan->pqs_filenamepath_maxlen = AOSegmentFilePathNameLen(relation) + 1;
scan->pqs_filenamepath = (char*)palloc0(scan->pqs_filenamepath_maxlen);
scan->pqs_rd = relation;
scan->parquetScanInitContext = CurrentMemoryContext;
/*
* Fill in Parquet Storage layer attributes.
*/
attr = &scan->storageAttributes;
/*
* These attributes describe the AppendOnly format to be scanned.
*/
if (aoEntry->compresstype == NULL || pg_strcasecmp(aoEntry->compresstype, "none") == 0)
attr->compress = false;
else
attr->compress = true;
if (aoEntry->compresstype != NULL)
attr->compressType = aoEntry->compresstype;
else
attr->compressType = "none";
attr->compressLevel = aoEntry->compresslevel;
attr->checksum = aoEntry->checksum;
attr->safeFSWriteSize = aoEntry->safefswritesize;
attr->splitsize = aoEntry->splitsize;
attr->version = aoEntry->version;
AORelationVersion_CheckValid(attr->version);
scan->proj = proj;
scan->pqs_tupDesc = (relationTupleDesc == NULL) ? RelationGetDescr(relation) : relationTupleDesc;
scan->hawqAttrToParquetColChunks = (int*)palloc0(scan->pqs_tupDesc->natts * sizeof(int));
initscan(scan);
return scan ;
}
