greenplumn extaccess 源码
greenplumn extaccess 代码
文件路径：/gpcontrib/gp_exttable_fdw/extaccess.c
/*-------------------------------------------------------------------------
*
* extaccess.c
* external table access method routines
*
* This access layer mimics the heap access API with respect to how it
* communicates with its respective scan node (external scan node) but
* instead of accessing the heap pages, it actually "scans" data by
* reading it from a local flat file or a remote data source.
*
* The actual data access, whether local or remote, is done with the
* curl c library ('libcurl') which uses a 'c-file like' API but behind
* the scenes actually does all the work of parsing the URI and communicating
* with the target. In this case if the URI uses the file protocol (file://)
* curl will try to open the specified file locally. If the URI uses the
* http protocol (http://) then curl will reach out to that address and
* get the data from there.
*
* As data is being read it gets parsed with the COPY command parsing rules,
* as if it is data meant for COPY. Therefore, currently, with the lack of
* single row error handling the first error will raise an error and the
* query will terminate.
*
* Portions Copyright (c) 2007-2008, Greenplum inc
* Portions Copyright (c) 2012-Present VMware, Inc. or its affiliates.
*
*
* IDENTIFICATION
* gpcontrib/gp_exttable_fdw/extaccess.c
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
#include "extaccess.h"
#include "access/external.h"
#include "access/formatter.h"
#include "access/heapam.h"
#include "access/url.h"
#include "access/valid.h"
#include "catalog/pg_proc.h"
#include "cdb/cdbsreh.h"
#include "cdb/cdbutil.h"
#include "cdb/cdbvars.h"
#include "commands/copy.h"
#include "commands/defrem.h"
#include "funcapi.h"
#include "mb/pg_wchar.h"
#include "nodes/makefuncs.h"
#include "pgstat.h"
#include "parser/parse_func.h"
#include "utils/relcache.h"
#include "utils/lsyscache.h"
#include "utils/memutils.h"
static HeapTuple externalgettup(FileScanDesc scan, ScanDirection dir);
static void InitParseState(CopyState pstate, Relation relation,
bool writable,
char fmtType,
char *uri, int rejectlimit,
bool islimitinrows, char logerrors);
static void FunctionCallPrepareFormatter(FunctionCallInfoBaseData *fcinfo,
int nArgs,
CopyState pstate,
FmgrInfo *formatter_func,
List *formatter_params,
FormatterData *formatter,
Relation rel,
TupleDesc tupDesc,
FmgrInfo *convFuncs,
Oid *typioparams);
static void open_external_readable_source(FileScanDesc scan, ExternalSelectDesc desc);
static void open_external_writable_source(ExternalInsertDesc extInsertDesc);
static int external_getdata_callback(void *outbuf, int minread, int maxread, void *extra);
static int external_getdata(URL_FILE *extfile, CopyState pstate, void *outbuf, int maxread);
static void external_senddata(URL_FILE *extfile, CopyState pstate);
static void external_scan_error_callback(void *arg);
static Oid lookupCustomFormatter(List **options, bool iswritable);
static void justifyDatabuf(StringInfo buf);
/* ----------------------------------------------------------------
* external_ interface functions
* ----------------------------------------------------------------
*/
#ifdef FILEDEBUGALL
#define FILEDEBUG_1 \
elog(DEBUG2, "external_getnext([%s],dir=%d) called", \
RelationGetRelationName(scan->fs_rd), (int) direction)
#define FILEDEBUG_2 \
elog(DEBUG2, "external_getnext returning EOS")
#define FILEDEBUG_3 \
elog(DEBUG2, "external_getnext returning tuple")
#else
#define FILEDEBUG_1
#define FILEDEBUG_2
#define FILEDEBUG_3
#endif /* !defined(FILEDEBUGALL) */
/* ----------------
* external_beginscan - begin file scan
* ----------------
*/
FileScanDesc
external_beginscan(Relation relation, uint32 scancounter,
List *uriList, char fmtType, bool isMasterOnly,
int rejLimit, bool rejLimitInRows, char logErrors, int encoding,
List *extOptions)
{
FileScanDesc scan;
TupleDesc tupDesc = NULL;
int attnum;
int segindex = GpIdentity.segindex;
char *uri = NULL;
List *custom_formatter_params = NIL;
/*
* 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 and initialize scan descriptor
*/
scan = (FileScanDesc) palloc0(sizeof(FileScanDescData));
scan->fs_ctup.t_data = NULL;
ItemPointerSetInvalid(&scan->fs_ctup.t_self);
scan->fs_rd = relation;
scan->fs_scancounter = scancounter;
scan->fs_noop = false;
scan->fs_file = NULL;
scan->fs_formatter = NULL;
scan->fs_constraintExprs = NULL;
if (relation->rd_att->constr != NULL && relation->rd_att->constr->num_check > 0)
scan->fs_hasConstraints = true;
else
scan->fs_hasConstraints = false;
scan->fs_isPartition = relation->rd_rel->relispartition;
/*
* get the external URI assigned to us.
*
* The URI assigned for this segment is normally in the uriList list at
* the index of this segment id. However, if we are executing on COORDINATOR
* ONLY the (one and only) entry which is destined for the master will be
* at the first entry of the uriList list.
*/
if (Gp_role == GP_ROLE_EXECUTE)
{
/* this is the normal path for most ext tables */
Value *v;
int idx = segindex;
/*
* Segindex may be -1, for the following case. A slice is executed on
* entry db, (for example, gp_segment_configuration), then external table is
* executed on another slice. Entry db slice will still call
* ExecInitExternalScan (probably we should fix this?), then segindex
* = -1 will bomb out here.
*/
if (isMasterOnly && idx == -1)
idx = 0;
if (idx >= 0)
{
v = (Value *) list_nth(uriList, idx);
if (v->type == T_Null)
uri = NULL;
else
uri = (char *) strVal(v);
}
}
else if (Gp_role == GP_ROLE_DISPATCH && isMasterOnly)
{
/* this is a ON COORDINATOR table. Only get uri if we are the master */
if (segindex == -1)
{
Value *v = list_nth(uriList, 0);
if (v->type == T_Null)
uri = NULL;
else
uri = (char *) strVal(v);
}
}
/*
* if a uri is assigned to us - get a reference to it. Some executors
* don't have a uri to scan (if # of uri's < # of primary segdbs). in
* which case uri will be NULL. If that's the case for this segdb set to
* no-op.
*/
if (uri)
{
/* set external source (uri) */
scan->fs_uri = uri;
/*
* NOTE: we delay actually opening the data source until
* external_getnext()
*/
}
else
{
/* segdb has no work to do. set to no-op */
scan->fs_noop = true;
scan->fs_uri = NULL;
}
tupDesc = RelationGetDescr(relation);
scan->fs_tupDesc = tupDesc;
//scan->attr = tupDesc->attrs;
scan->num_phys_attrs = tupDesc->natts;
scan->values = (Datum *) palloc(scan->num_phys_attrs * sizeof(Datum));
scan->nulls = (bool *) palloc(scan->num_phys_attrs * sizeof(bool));
/*
* Pick up the required catalog information for each attribute in the
* relation, including the input function and the element type (to pass to
* the input function).
*/
scan->in_functions = (FmgrInfo *) palloc(scan->num_phys_attrs * sizeof(FmgrInfo));
scan->typioparams = (Oid *) palloc(scan->num_phys_attrs * sizeof(Oid));
for (attnum = 1; attnum <= scan->num_phys_attrs; attnum++)
{
/* We don't need info for dropped attributes */
Form_pg_attribute attr = TupleDescAttr(scan->fs_tupDesc, attnum - 1);
if (attr->attisdropped)
continue;
getTypeInputInfo(attr->atttypid,
&scan->in_func_oid, &scan->typioparams[attnum - 1]);
fmgr_info(scan->in_func_oid, &scan->in_functions[attnum - 1]);
}
custom_formatter_params = list_copy(extOptions);
/*
* pass external table's encoding to copy's options
*
* don't append to entry->options directly, we only store the encoding in
* entry->encoding (and ftoptions)
*/
extOptions = appendCopyEncodingOption(list_copy(extOptions), encoding);
/*
* Allocate and init our structure that keeps track of data parsing state
*/
scan->fs_pstate = BeginCopyFrom(NULL,
relation, NULL, false,
external_getdata_callback,
(void *) scan,
NIL,
(fmttype_is_custom(fmtType) ? NIL : extOptions));
if (scan->fs_pstate->header_line && Gp_role == GP_ROLE_DISPATCH)
{
ereport(NOTICE,
(errmsg("HEADER means that each one of the data files has a header row")));
}
/* Initialize all the parsing and state variables */
InitParseState(scan->fs_pstate, relation, false, fmtType,
scan->fs_uri, rejLimit, rejLimitInRows, logErrors);
if (fmttype_is_custom(fmtType))
{
/*
* Custom format: get formatter name and find it in the catalog
*/
Oid procOid;
/* parseFormatString should have seen a formatter name */
procOid = lookupCustomFormatter(&custom_formatter_params, false);
/* we found our function. set it up for calling */
scan->fs_custom_formatter_func = palloc(sizeof(FmgrInfo));
fmgr_info(procOid, scan->fs_custom_formatter_func);
scan->fs_custom_formatter_params = custom_formatter_params;
scan->fs_formatter = (FormatterData *) palloc0(sizeof(FormatterData));
initStringInfo(&scan->fs_formatter->fmt_databuf);
scan->fs_formatter->fmt_perrow_ctx = scan->fs_pstate->rowcontext;
}
/* pgstat_initstats(relation); */
return scan;
}
/* ----------------
* external_rescan - (re)start a scan of an external file
* ----------------
*/
void
external_rescan(FileScanDesc scan)
{
/* Close previous scan if it was already open */
external_stopscan(scan);
/* The first call to external_getnext will re-open the scan */
if (!scan->fs_pstate)
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("The file parse state of external scan is invalid")));
/* reset some parse state variables */
scan->fs_pstate->reached_eof = false;
scan->fs_pstate->cur_lineno = 0;
scan->fs_pstate->cur_attname = NULL;
scan->fs_pstate->raw_buf_len = 0;
}
/* ----------------
* external_endscan - end a scan
* ----------------
*/
void
external_endscan(FileScanDesc scan)
{
char *relname = pstrdup(RelationGetRelationName(scan->fs_rd));
if (scan->fs_pstate != NULL)
{
/*
* decrement relation reference count and free scan descriptor storage
*/
RelationDecrementReferenceCount(scan->fs_rd);
}
if (scan->values)
{
pfree(scan->values);
scan->values = NULL;
}
if (scan->nulls)
{
pfree(scan->nulls);
scan->nulls = NULL;
}
if (scan->in_functions)
{
pfree(scan->in_functions);
scan->in_functions = NULL;
}
if (scan->typioparams)
{
pfree(scan->typioparams);
scan->typioparams = NULL;
}
if (scan->fs_pstate != NULL && scan->fs_pstate->rowcontext != NULL)
{
/*
* delete the row context
*/
MemoryContextDelete(scan->fs_pstate->rowcontext);
scan->fs_pstate->rowcontext = NULL;
}
/*----
* if SREH was active:
* 1) QEs: send a libpq message to QD with num of rows rejected in this segment
* 2) Free SREH resources
*----
*/
if (scan->fs_pstate != NULL && scan->fs_pstate->errMode != ALL_OR_NOTHING)
{
if (Gp_role == GP_ROLE_EXECUTE)
SendNumRows(scan->fs_pstate->cdbsreh->rejectcount, 0);
destroyCdbSreh(scan->fs_pstate->cdbsreh);
}
if (scan->fs_formatter)
{
/*
* TODO: check if this space is automatically freed. if not, then see
* what about freeing the user context
*/
if (scan->fs_formatter->fmt_databuf.data)
pfree(scan->fs_formatter->fmt_databuf.data);
pfree(scan->fs_formatter);
scan->fs_formatter = NULL;
}
/*
* free parse state memory
*/
if (scan->fs_pstate != NULL)
{
if (scan->fs_pstate->attribute_buf.data)
pfree(scan->fs_pstate->attribute_buf.data);
if (scan->fs_pstate->line_buf.data)
pfree(scan->fs_pstate->line_buf.data);
if (scan->fs_pstate->force_quote_flags)
pfree(scan->fs_pstate->force_quote_flags);
if (scan->fs_pstate->force_notnull_flags)
pfree(scan->fs_pstate->force_notnull_flags);
pfree(scan->fs_pstate);
scan->fs_pstate = NULL;
}
/*
* Close the external file
*/
if (!scan->fs_noop && scan->fs_file)
{
/*
* QueryFinishPending == true means QD have got
* enough tuples and query can return correctly,
* so slient errors when closing external file.
*/
url_fclose(scan->fs_file, !QueryFinishPending, relname);
scan->fs_file = NULL;
}
pfree(relname);
}
/* ----------------
* external_stopscan - closes an external resource without dismantling the scan context
* ----------------
*/
void
external_stopscan(FileScanDesc scan)
{
/*
* Close the external file
*/
if (!scan->fs_noop && scan->fs_file)
{
url_fclose(scan->fs_file, false, RelationGetRelationName(scan->fs_rd));
scan->fs_file = NULL;
}
}
/* ----------------
* external_getnext_init - prepare ExternalSelectDesc struct before external_getnext
* ----------------
*/
ExternalSelectDesc
external_getnext_init(PlanState *state)
{
ExternalSelectDesc
desc = (ExternalSelectDesc) palloc0(sizeof(ExternalSelectDescData));
if (state != NULL)
desc->projInfo = state->ps_ProjInfo;
return desc;
}
/* ----------------------------------------------------------------
* external_getnext
*
* Parse a data file and return its rows in heap tuple form
* ----------------------------------------------------------------
*/
HeapTuple
external_getnext(FileScanDesc scan, ScanDirection direction, ExternalSelectDesc desc)
{
HeapTuple tuple;
if (scan->fs_noop)
return NULL;
/*
* open the external source (local file or http).
*
* NOTE: external_beginscan() seems like the natural place for this call.
* However, in queries with more than one gang each gang will initialized
* all the nodes of the plan (but actually executed only the nodes in it's
* local slice) This means that external_beginscan() (and
* external_endscan() too) will get called more than needed and we'll end
* up opening too many http connections when they are not expected (see
* MPP-1261). Therefore we instead do it here on the first time around
* only.
*/
if (!scan->fs_file)
open_external_readable_source(scan, desc);
/* Note: no locking manipulations needed */
FILEDEBUG_1;
tuple = externalgettup(scan, direction);
if (tuple == NULL)
{
FILEDEBUG_2; /* external_getnext returning EOS */
return NULL;
}
/*
* if we get here it means we have a new current scan tuple
*/
FILEDEBUG_3; /* external_getnext returning tuple */
pgstat_count_heap_getnext(scan->fs_rd);
return tuple;
}
/*
* external_insert_init
*
* before using external_insert() to insert tuples we need to call
* this function to initialize our various structures and state..
*/
ExternalInsertDesc
external_insert_init(Relation rel)
{
ExternalInsertDesc extInsertDesc;
ExtTableEntry *extentry;
List *copyFmtOpts;
List *custom_formatter_params = NIL;
char *on_clause;
/*
* Get the ExtTableEntry information for this table
*/
extentry = GetExtTableEntry(RelationGetRelid(rel));
on_clause = (char *) strVal(linitial(extentry->execlocations));
/*
* allocate and initialize the insert descriptor
*/
extInsertDesc = (ExternalInsertDesc) palloc0(sizeof(ExternalInsertDescData));
extInsertDesc->ext_rel = rel;
if (strcmp(on_clause, "COORDINATOR_ONLY") == 0)
extInsertDesc->ext_noop = false;
else
extInsertDesc->ext_noop = (Gp_role == GP_ROLE_DISPATCH);
extInsertDesc->ext_formatter_data = NULL;
if (extentry->command)
{
/*
* EXECUTE
*
* build the command string, 'execute:<command>'
*/
extInsertDesc->ext_uri = psprintf("execute:%s", extentry->command);
}
else
{
/* LOCATION - gpfdist or custom */
Value *v;
char *uri_str;
int segindex = GpIdentity.segindex;
int num_segs = getgpsegmentCount();
int num_urls = list_length(extentry->urilocations);
int my_url = segindex % num_urls;
if (num_urls > num_segs)
ereport(ERROR,
(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
errmsg("external table has more URLs than available primary segments that can write into them")));
/* get a url to use. we use seg number modulo total num of urls */
v = list_nth(extentry->urilocations, my_url);
uri_str = pstrdup(v->val.str);
extInsertDesc->ext_uri = uri_str;
#if 0
elog(NOTICE, "seg %d got url number %d: %s", segindex, my_url, uri_str);
#endif
}
/*
* Allocate and init our structure that keeps track of data parsing state
*/
extInsertDesc->ext_pstate = (CopyStateData *) palloc0(sizeof(CopyStateData));
extInsertDesc->ext_tupDesc = RelationGetDescr(rel);
/*
* Writing to an external table is like COPY TO: we get tuples from the executor,
* we format them into the format requested, and write the output to an external
* sink.
*/
custom_formatter_params = list_copy(extentry->options);
/*
* pass external table's encoding to copy's options
*
* don't append to entry->options directly, we only store the encoding in
* entry->encoding (and ftoptions)
*/
copyFmtOpts = appendCopyEncodingOption(list_copy(extentry->options), extentry->encoding);
extInsertDesc->ext_pstate = BeginCopyToForeignTable(rel, (fmttype_is_custom(extentry->fmtcode) ? NIL : copyFmtOpts));
InitParseState(extInsertDesc->ext_pstate,
rel,
true,
extentry->fmtcode,
extInsertDesc->ext_uri,
extentry->rejectlimit,
(extentry->rejectlimittype == 'r'),
extentry->logerrors);
if (fmttype_is_custom(extentry->fmtcode))
{
/*
* Custom format: get formatter name and find it in the catalog
*/
Oid procOid;
procOid = lookupCustomFormatter(&custom_formatter_params, true);
/* we found our function. set it up for calling */
extInsertDesc->ext_custom_formatter_func = palloc(sizeof(FmgrInfo));
fmgr_info(procOid, extInsertDesc->ext_custom_formatter_func);
extInsertDesc->ext_custom_formatter_params = custom_formatter_params;
extInsertDesc->ext_formatter_data = (FormatterData *) palloc0(sizeof(FormatterData));
extInsertDesc->ext_formatter_data->fmt_perrow_ctx = extInsertDesc->ext_pstate->rowcontext;
}
return extInsertDesc;
}
/*
* external_insert - format the tuple into text and write to the external source
*
* Note the following major differences from heap_insert
*
* - wal is always bypassed here.
* - transaction information is of no interest.
* - tuples are sent always to the destination (local file or remote target).
*
* Like heap_insert(), this function can modify the input tuple!
*/
void
external_insert(ExternalInsertDesc extInsertDesc, TupleTableSlot *slot)
{
TupleDesc tupDesc = extInsertDesc->ext_tupDesc;
CopyStateData *pstate = extInsertDesc->ext_pstate;
bool customFormat = (extInsertDesc->ext_custom_formatter_func != NULL);
if (extInsertDesc->ext_noop)
return;
/* Open our output file or output stream if not yet open */
if (!extInsertDesc->ext_file && !extInsertDesc->ext_noop)
open_external_writable_source(extInsertDesc);
/*
* deconstruct the tuple and format it into text
*/
if (!customFormat)
{
/* TEXT or CSV */
CopyOneRowTo(pstate, slot);
CopySendEndOfRow(pstate);
}
else
{
/* custom format. convert tuple using user formatter */
Datum d;
bytea *b;
LOCAL_FCINFO(fcinfo, 1);
HeapTuple instup;
/*
* There is some redundancy between FormatterData and
* ExternalInsertDesc we may be able to consolidate data structures a
* little.
*/
FormatterData *formatter = extInsertDesc->ext_formatter_data;
/* must have been created during insert_init */
Assert(formatter);
/* per call formatter prep */
FunctionCallPrepareFormatter(fcinfo,
1,
pstate,
extInsertDesc->ext_custom_formatter_func,
extInsertDesc->ext_custom_formatter_params,
formatter,
extInsertDesc->ext_rel,
extInsertDesc->ext_tupDesc,
pstate->out_functions,
NULL);
/* Mark the correct record type in the passed tuple */
/*
* get the heap tuple out of the tuple table slot, making sure we have a
* writable copy. (the function can scribble on the tuple)
*/
instup = ExecCopySlotHeapTuple(slot);
HeapTupleHeaderSetDatumLength(instup->t_data, instup->t_len);
HeapTupleHeaderSetTypeId(instup->t_data, tupDesc->tdtypeid);
HeapTupleHeaderSetTypMod(instup->t_data, tupDesc->tdtypmod);
fcinfo->args[0].value = HeapTupleGetDatum(instup);
fcinfo->args[0].isnull = false;
d = FunctionCallInvoke(fcinfo);
MemoryContextReset(formatter->fmt_perrow_ctx);
/* We do not expect a null result */
if (fcinfo->isnull)
elog(ERROR, "function %u returned NULL", fcinfo->flinfo->fn_oid);
b = DatumGetByteaP(d);
CopyOneCustomRowTo(pstate, b);
heap_freetuple(instup);
}
/* Write the data into the external source */
external_senddata(extInsertDesc->ext_file, pstate);
/* Reset our buffer to start clean next round */
pstate->fe_msgbuf->len = 0;
pstate->fe_msgbuf->data[0] = '\0';
}
/*
* external_insert_finish
*
* when done inserting all the data via external_insert() we need to call
* this function to flush all remaining data in the buffer into the file.
*/
void
external_insert_finish(ExternalInsertDesc extInsertDesc)
{
/*
* Close the external source
*/
if (extInsertDesc->ext_file)
{
char *relname = RelationGetRelationName(extInsertDesc->ext_rel);
url_fflush(extInsertDesc->ext_file, extInsertDesc->ext_pstate);
url_fclose(extInsertDesc->ext_file, true, relname);
}
if (extInsertDesc->ext_formatter_data)
pfree(extInsertDesc->ext_formatter_data);
pfree(extInsertDesc);
}
/* ==========================================================================
* The follwing macros aid in major refactoring of data processing code (in
* externalgettup() ). We use macros because in some cases the code must be in
* line in order to work (for example elog_dismiss() in PG_CATCH) while in
* other cases we'd like to inline the code for performance reasons.
*
* NOTE that an almost identical set of macros exists in copy.c for the COPY
* command. If you make changes here you may want to consider taking a look.
* ==========================================================================
*/
#define EXT_RESET_LINEBUF \
pstate->line_buf.len = 0; \
pstate->line_buf.data[0] = '\0'; \
pstate->line_buf.cursor = 0;
/*
* A data error happened. This code block will always be inside a PG_CATCH()
* block right when a higher stack level produced an error. We handle the error
* by checking which error mode is set (SREH or all-or-nothing) and do the right
* thing accordingly. Note that we MUST have this code in a macro (as opposed
* to a function) as elog_dismiss() has to be inlined with PG_CATCH in order to
* access local error state variables.
*/
#define FILEAM_HANDLE_ERROR \
if (pstate->errMode == ALL_OR_NOTHING) \
{ \
/* re-throw error and abort */ \
PG_RE_THROW(); \
} \
else \
{ \
/* SREH - release error state */ \
\
ErrorData *edata; \
MemoryContext oldcontext;\
\
/* SREH must only handle data errors. all other errors must not be caught */\
if(ERRCODE_TO_CATEGORY(elog_geterrcode()) != ERRCODE_DATA_EXCEPTION)\
{\
PG_RE_THROW(); \
}\
\
/* save a copy of the error info */ \
oldcontext = MemoryContextSwitchTo(pstate->cdbsreh->badrowcontext);\
edata = CopyErrorData();\
MemoryContextSwitchTo(oldcontext);\
\
if (!elog_dismiss(DEBUG5)) \
PG_RE_THROW(); /* <-- hope to never get here! */ \
\
truncateEol(&pstate->line_buf, pstate->eol_type); \
pstate->cdbsreh->rawdata->cursor = 0; \
pstate->cdbsreh->rawdata->data = pstate->line_buf.data; \
pstate->cdbsreh->rawdata->len = pstate->line_buf.len; \
pstate->cdbsreh->is_server_enc = pstate->line_buf_converted; \
pstate->cdbsreh->linenumber = pstate->cur_lineno; \
pstate->cdbsreh->processed++; \
\
/* set the error message. Use original msg and add column name if available */ \
if (pstate->cur_attname)\
{\
pstate->cdbsreh->errmsg = psprintf("%s, column %s", \
edata->message, \
pstate->cur_attname); \
}\
else\
{\
pstate->cdbsreh->errmsg = pstrdup(edata->message); \
}\
\
HandleSingleRowError(pstate->cdbsreh); \
FreeErrorData(edata);\
if (!IsRejectLimitReached(pstate->cdbsreh)) \
pfree(pstate->cdbsreh->errmsg); \
}
static HeapTuple
externalgettup_defined(FileScanDesc scan)
{
HeapTuple tuple = NULL;
CopyState pstate = scan->fs_pstate;
MemoryContext oldcontext;
MemoryContextReset(pstate->rowcontext);
oldcontext = MemoryContextSwitchTo(pstate->rowcontext);
/* Get a line */
if (!NextCopyFrom(pstate,
NULL,
scan->values,
scan->nulls))
{
MemoryContextSwitchTo(oldcontext);
return NULL;
}
MemoryContextSwitchTo(oldcontext);
/* convert to heap tuple */
/*
* XXX This is bad code. Planner should be able to decide
* whether we need heaptuple or memtuple upstream, so make the
* right decision here.
*/
tuple = heap_form_tuple(scan->fs_tupDesc, scan->values, scan->nulls);
return tuple;
}
static HeapTuple
externalgettup_custom(FileScanDesc scan)
{
HeapTuple tuple;
CopyState pstate = scan->fs_pstate;
FormatterData *formatter = scan->fs_formatter;
MemoryContext oldctxt = CurrentMemoryContext;
Assert(formatter);
Assert(pstate->raw_buf_len >= 0);
/* while didn't finish processing the entire file */
/* raw_buf_len was set to 0 in BeginCopyFrom() or external_rescan() */
while (pstate->raw_buf_len != 0 || !pstate->reached_eof)
{
/* need to fill our buffer with data? */
if (pstate->raw_buf_len == 0)
{
int bytesread = external_getdata(scan->fs_file, pstate, pstate->raw_buf, RAW_BUF_SIZE);
if (bytesread > 0)
{
appendBinaryStringInfo(&formatter->fmt_databuf, pstate->raw_buf, bytesread);
pstate->raw_buf_len = bytesread;
}
/* HEADER not yet supported ... */
if (pstate->header_line)
elog(ERROR, "header line in custom format is not yet supported");
}
/* while there is still data in our buffer */
while (pstate->raw_buf_len > 0)
{
bool error_caught = false;
/*
* Invoke the custom formatter function.
*/
PG_TRY();
{
LOCAL_FCINFO(fcinfo, 0);
/* per call formatter prep */
FunctionCallPrepareFormatter(fcinfo,
0,
pstate,
scan->fs_custom_formatter_func,
scan->fs_custom_formatter_params,
formatter,
scan->fs_rd,
scan->fs_tupDesc,
scan->in_functions,
scan->typioparams);
(void) FunctionCallInvoke(fcinfo);
}
PG_CATCH();
{
error_caught = true;
MemoryContextSwitchTo(formatter->fmt_perrow_ctx);
/*
* Save any bad row information that was set by the user
* in the formatter UDF (if any). Then handle the error in
* FILEAM_HANDLE_ERROR.
*/
pstate->cur_lineno = formatter->fmt_badrow_num;
resetStringInfo(&pstate->line_buf);
if (formatter->fmt_badrow_len > 0)
{
if (formatter->fmt_badrow_data)
appendBinaryStringInfo(&pstate->line_buf,
formatter->fmt_badrow_data,
formatter->fmt_badrow_len);
formatter->fmt_databuf.cursor += formatter->fmt_badrow_len;
if (formatter->fmt_databuf.cursor > formatter->fmt_databuf.len ||
formatter->fmt_databuf.cursor < 0)
{
formatter->fmt_databuf.cursor = formatter->fmt_databuf.len;
}
}
FILEAM_HANDLE_ERROR;
FlushErrorState();
MemoryContextSwitchTo(oldctxt);
}
PG_END_TRY();
/*
* Examine the function results. If an error was caught we
* already handled it, so after checking the reject limit,
* loop again and call the UDF for the next tuple.
*/
if (!error_caught)
{
switch (formatter->fmt_notification)
{
case FMT_NONE:
/* got a tuple back */
tuple = formatter->fmt_tuple;
if (pstate->cdbsreh)
pstate->cdbsreh->processed++;
MemoryContextReset(formatter->fmt_perrow_ctx);
return tuple;
case FMT_NEED_MORE_DATA:
/*
* Callee consumed all data in the buffer. Prepare
* to read more data into it.
*/
pstate->raw_buf_len = 0;
justifyDatabuf(&formatter->fmt_databuf);
continue;
default:
elog(ERROR, "unsupported formatter notification (%d)",
formatter->fmt_notification);
break;
}
}
else
{
ErrorIfRejectLimitReached(pstate->cdbsreh);
}
}
}
if (formatter->fmt_databuf.len > 0)
{
/*
* The formatter needs more data, but we have reached
* EOF. This is an error.
*/
ereport(WARNING,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("unexpected end of file")));
}
/*
* if we got here we finished reading all the data.
*/
return NULL;
}
/* ----------------
* externalgettup form another tuple from the data file.
* This is the workhorse - make sure it's fast!
*
* Initialize the scan if not already done.
* Verify that we are scanning forward only.
*
* ----------------
*/
static HeapTuple
externalgettup(FileScanDesc scan,
ScanDirection dir pg_attribute_unused())
{
bool custom = (scan->fs_custom_formatter_func != NULL);
HeapTuple tup = NULL;
ErrorContextCallback externalscan_error_context;
Assert(ScanDirectionIsForward(dir));
externalscan_error_context.callback = external_scan_error_callback;
externalscan_error_context.arg = (void *) scan;
externalscan_error_context.previous = error_context_stack;
error_context_stack = &externalscan_error_context;
if (!custom)
tup = externalgettup_defined(scan); /* text/csv */
else
tup = externalgettup_custom(scan); /* custom */
/* Restore the previous error callback */
error_context_stack = externalscan_error_context.previous;
return tup;
}
/*
* setCustomFormatter
*
* Given a formatter name and a function signature (pre-determined
* by whether it is readable or writable) find such a function in
* the catalog and store it to be used later.
*
* WET function: 1 record arg, return bytea.
* RET function: 0 args, returns record.
*/
static Oid
lookupCustomFormatter(List **options, bool iswritable)
{
ListCell *cell;
ListCell *prev = NULL;
char *formatter_name;
List *funcname = NIL;
Oid procOid = InvalidOid;
Oid argList[1];
Oid returnOid;
/*
* The formatter is defined as a "formatter=<name>" tuple in the options
* array. Extract into a separate list in order to scan the catalog for
* the function definition.
*/
foreach(cell, *options)
{
DefElem *defel = (DefElem *) lfirst(cell);
if (strcmp(defel->defname, "formatter") == 0)
{
formatter_name = defGetString(defel);
funcname = list_make1(makeString(formatter_name));
*options = list_delete_cell(*options, cell, prev);
break;
}
prev = cell;
}
if (list_length(funcname) == 0)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("formatter function not found in table options")));
if (iswritable)
{
argList[0] = RECORDOID;
returnOid = BYTEAOID;
procOid = LookupFuncName(funcname, 1, argList, true);
}
else
{
returnOid = RECORDOID;
procOid = LookupFuncName(funcname, 0, argList, true);
}
if (!OidIsValid(procOid))
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("formatter function \"%s\" of type %s was not found",
formatter_name,
(iswritable ? "writable" : "readable")),
errhint("Create it with CREATE FUNCTION.")));
/* check return type matches */
if (get_func_rettype(procOid) != returnOid)
ereport(ERROR,
(errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
errmsg("formatter function \"%s\" of type %s has an incorrect return type",
formatter_name,
(iswritable ? "writable" : "readable"))));
/* check allowed volatility */
if (func_volatile(procOid) != PROVOLATILE_STABLE)
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("formatter function %s is not declared STABLE",
formatter_name)));
return procOid;
}
/*
* Initialize the data parsing state.
*
* This includes format descriptions (delimiter, quote...), format type
* (text, csv), etc...
*/
static void
InitParseState(CopyState pstate, Relation relation,
bool iswritable,
char fmtType,
char *uri, int rejectlimit,
bool islimitinrows, char logerrors)
{
/*
* Error handling setup
*/
if (rejectlimit == -1)
{
/* Default error handling - "all-or-nothing" */
pstate->cdbsreh = NULL; /* no SREH */
pstate->errMode = ALL_OR_NOTHING;
}
else
{
/* select the SREH mode */
if (IS_LOG_TO_FILE(logerrors))
{
/* errors into file */
pstate->errMode = SREH_LOG;
}
else
{
/* no error log */
pstate->errMode = SREH_IGNORE;
}
/* Single row error handling */
pstate->cdbsreh = makeCdbSreh(rejectlimit,
islimitinrows,
uri,
(char *) pstate->cur_relname,
logerrors);
pstate->cdbsreh->relid = RelationGetRelid(relation);
}
/* Initialize 'out_functions', like CopyTo() would. */
CopyState cstate = pstate;
TupleDesc tupDesc = RelationGetDescr(cstate->rel);
int num_phys_attrs = tupDesc->natts;
cstate->out_functions = (FmgrInfo *) palloc(num_phys_attrs * sizeof(FmgrInfo));
ListCell *cur;
foreach(cur, cstate->attnumlist)
{
int attnum = lfirst_int(cur);
Form_pg_attribute attr = TupleDescAttr(tupDesc, attnum - 1);
Oid out_func_oid;
bool isvarlena;
if (cstate->binary)
getTypeBinaryOutputInfo(attr->atttypid,
&out_func_oid,
&isvarlena);
else
getTypeOutputInfo(attr->atttypid,
&out_func_oid,
&isvarlena);
fmgr_info(out_func_oid, &cstate->out_functions[attnum - 1]);
}
/* and 'fe_mgbuf' */
cstate->fe_msgbuf = makeStringInfo();
/*
* Create a temporary memory context that we can reset once per row to
* recover palloc'd memory. This avoids any problems with leaks inside
* datatype input or output routines, and should be faster than retail
* pfree's anyway.
*/
pstate->rowcontext = AllocSetContextCreate(CurrentMemoryContext,
"ExtTableMemCxt",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
}
/*
* Prepare the formatter data to be used inside the formatting UDF.
* This function should be called every time before invoking the
* UDF, for both insert and scan operations. Even though there's some
* redundancy here, it is needed in order to reset some per-call state
* that should be examined by the caller upon return from the UDF.
*
* Also, set up the function call context.
*/
static void
FunctionCallPrepareFormatter(FunctionCallInfoBaseData *fcinfo,
int nArgs,
CopyState pstate,
FmgrInfo *formatter_func,
List *formatter_params,
FormatterData *formatter,
Relation rel,
TupleDesc tupDesc,
FmgrInfo *convFuncs,
Oid *typioparams)
{
formatter->type = T_FormatterData;
formatter->fmt_relation = rel;
formatter->fmt_tupDesc = tupDesc;
formatter->fmt_notification = FMT_NONE;
formatter->fmt_badrow_len = 0;
formatter->fmt_badrow_num = 0;
formatter->fmt_args = formatter_params;
formatter->fmt_conv_funcs = convFuncs;
formatter->fmt_saw_eof = pstate->reached_eof;
formatter->fmt_typioparams = typioparams;
formatter->fmt_perrow_ctx = pstate->rowcontext;
formatter->fmt_needs_transcoding = pstate->need_transcoding;
formatter->fmt_conversion_proc = pstate->enc_conversion_proc;
formatter->fmt_external_encoding = pstate->file_encoding;
InitFunctionCallInfoData( /* FunctionCallInfoData */ *fcinfo,
/* FmgrInfo */ formatter_func,
/* nArgs */ nArgs,
/* collation */ InvalidOid,
/* Call Context */ (Node *) formatter,
/* ResultSetInfo */ NULL);
}
/*
* open the external source for scanning (RET only)
*
* an external source is one of the following:
* 1) a local file (requested by 'file')
* 2) a remote http server
* 3) a remote gpfdist server
* 4) a command to execute
*/
static void
open_external_readable_source(FileScanDesc scan, ExternalSelectDesc desc)
{
extvar_t extvar;
/* set up extvar */
memset(&extvar, 0, sizeof(extvar));
external_set_env_vars_ext(&extvar,
scan->fs_uri,
scan->fs_pstate->csv_mode,
scan->fs_pstate->escape,
scan->fs_pstate->quote,
scan->fs_pstate->eol_type,
scan->fs_pstate->header_line,
scan->fs_scancounter,
scan->fs_custom_formatter_params);
/* actually open the external source */
scan->fs_file = url_fopen(scan->fs_uri,
false /* for read */ ,
&extvar,
scan->fs_pstate,
desc);
}
/*
* open the external source for writing (WET only)
*
* an external source is one of the following:
* 1) a local file (requested by 'tablespace' protocol)
* 2) a command to execute
*/
static void
open_external_writable_source(ExternalInsertDesc extInsertDesc)
{
extvar_t extvar;
/* set up extvar */
memset(&extvar, 0, sizeof(extvar));
external_set_env_vars_ext(&extvar,
extInsertDesc->ext_uri,
extInsertDesc->ext_pstate->csv_mode,
extInsertDesc->ext_pstate->escape,
extInsertDesc->ext_pstate->quote,
extInsertDesc->ext_pstate->eol_type,
extInsertDesc->ext_pstate->header_line,
0,
extInsertDesc->ext_custom_formatter_params);
/* actually open the external source */
extInsertDesc->ext_file = url_fopen(extInsertDesc->ext_uri,
true /* forwrite */ ,
&extvar,
extInsertDesc->ext_pstate,
NULL);
}
/*
* Fetch more data from the source, and feed it to the COPY FROM machinery
* for parsing.
*/
static int
external_getdata_callback(void *outbuf, int minread, int maxread, void *extra)
{
FileScanDesc scan = (FileScanDesc) extra;
return external_getdata(scan->fs_file, scan->fs_pstate, outbuf, maxread);
}
/*
* get a chunk of data from the external data file.
*/
static int
external_getdata(URL_FILE *extfile, CopyState pstate, void *outbuf, int maxread)
{
int bytesread;
/*
* CK: this code is very delicate. The caller expects this: - if url_fread
* returns something, and the EOF is reached, it this call must return
* with both the content and the reached_eof flag set. - failing to do so will
* result in skipping the last line.
*/
bytesread = url_fread((void *) outbuf, maxread, extfile, pstate);
if (url_feof(extfile, bytesread))
{
pstate->reached_eof = true;
}
if (bytesread <= 0)
{
if (url_ferror(extfile, bytesread, NULL, 0))
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not read from external file: %m")));
}
return bytesread;
}
/*
* send a chunk of data from the external data file.
*/
static void
external_senddata(URL_FILE *extfile, CopyState pstate)
{
StringInfo fe_msgbuf = pstate->fe_msgbuf;
static char ebuf[512] = {0};
size_t nwrote = 0;
int ebuflen = 512;
nwrote = url_fwrite((void *) fe_msgbuf->data, fe_msgbuf->len, extfile, pstate);
if (url_ferror(extfile, nwrote, ebuf, ebuflen))
{
if (*ebuf && strlen(ebuf) > 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not write to external resource: %s",
ebuf)));
else
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not write to external resource: %m")));
}
}
static char *
linenumber_atoi(char *buffer, size_t bufsz, int64 linenumber)
{
if (linenumber < 0)
snprintf(buffer, bufsz, "%s", "N/A");
else
snprintf(buffer, bufsz, INT64_FORMAT, linenumber);
return buffer;
}
/*
* error context callback for external table scan
*/
static void
external_scan_error_callback(void *arg)
{
FileScanDesc scan = (FileScanDesc) arg;
CopyState cstate = scan->fs_pstate;
char buffer[20];
/*
* early exit for custom format error. We don't have metadata to report
* on. TODO: this actually will override any errcontext that the user
* wants to set. maybe another approach is needed here.
*/
if (scan->fs_custom_formatter_func)
{
errcontext("External table %s", cstate->cur_relname);
return;
}
if (cstate->cur_attname)
{
/* error is relevant to a particular column */
errcontext("External table %s, line %s of %s, column %s",
cstate->cur_relname,
linenumber_atoi(buffer, sizeof(buffer), cstate->cur_lineno),
scan->fs_uri,
cstate->cur_attname);
}
else
{
/* error is relevant to a particular line */
if (cstate->line_buf_converted || !cstate->need_transcoding)
{
char *line_buf;
line_buf = limit_printout_length(cstate->line_buf.data);
truncateEolStr(line_buf, cstate->eol_type);
errcontext("External table %s, line %s of %s: \"%s\"",
cstate->cur_relname,
linenumber_atoi(buffer, sizeof(buffer), cstate->cur_lineno),
scan->fs_uri, line_buf);
pfree(line_buf);
}
else
{
/*
* Here, the line buffer is still in a foreign encoding, and
* indeed it's quite likely that the error is precisely a failure
* to do encoding conversion (ie, bad data). We dare not try to
* convert it, and at present there's no way to regurgitate it
* without conversion. So we have to punt and just report the
* line number.
*
* since the gpfdist protocol does not transfer line numbers
* correclty in certain places - if line number is 0 we just do
* not print it.
*/
if (cstate->cur_lineno > 0)
errcontext("External table %s, line %s of file %s",
cstate->cur_relname,
linenumber_atoi(buffer, sizeof(buffer), cstate->cur_lineno),
scan->fs_uri);
else
errcontext("External table %s, file %s",
cstate->cur_relname, scan->fs_uri);
}
}
}
/*
* justifyDatabuf
*
* shift all data remaining in the buffer (anything from cursor to
* end of buffer) to the beginning of the buffer, and readjust the
* cursor and length to the new end of buffer position.
*
* 3 possible cases:
* 1 - cursor at beginning of buffer (whole buffer is a partial row) - nothing to do.
* 2 - cursor at end of buffer (last row ended in the last byte of the buffer)
* 3 - cursor at middle of buffer (remaining bytes are a partial row)
*/
static void
justifyDatabuf(StringInfo buf)
{
/* 1 */
if (buf->cursor == 0)
{
/* nothing to do */
}
/* 2 */
else if (buf->cursor == buf->len)
{
Assert(buf->data[buf->cursor] == '\0');
resetStringInfo(buf);
}
/* 3 */
else
{
char *position = buf->data + buf->cursor;
int remaining = buf->len - buf->cursor;
/* slide data back (data may overlap so use memmove not memcpy) */
memmove(buf->data, position, remaining);
buf->len = remaining;
buf->data[buf->len] = '\0'; /* be consistent with
* appendBinaryStringInfo() */
}
buf->cursor = 0;
}
List *
appendCopyEncodingOption(List *copyFmtOpts, int encoding)
{
return lappend(copyFmtOpts,
makeDefElem("encoding",
(Node *)makeString((char *)pg_encoding_to_char(encoding)),
-1));
}
相关信息
相关文章
0
赞
热门推荐
-
2、 - 优质文章
-
3、 gate.io
-
8、 golang
-
9、 openharmony
-
10、 Vue中input框自动聚焦