Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * copyfromparse.c
4 : * Parse CSV/text/binary format for COPY FROM.
5 : *
6 : * This file contains routines to parse the text, CSV and binary input
7 : * formats. The main entry point is NextCopyFrom(), which parses the
8 : * next input line and returns it as Datums.
9 : *
10 : * In text/CSV mode, the parsing happens in multiple stages:
11 : *
12 : * [data source] --> raw_buf --> input_buf --> line_buf --> attribute_buf
13 : * 1. 2. 3. 4.
14 : *
15 : * 1. CopyLoadRawBuf() reads raw data from the input file or client, and
16 : * places it into 'raw_buf'.
17 : *
18 : * 2. CopyConvertBuf() calls the encoding conversion function to convert
19 : * the data in 'raw_buf' from client to server encoding, placing the
20 : * converted result in 'input_buf'.
21 : *
22 : * 3. CopyReadLine() parses the data in 'input_buf', one line at a time.
23 : * It is responsible for finding the next newline marker, taking quote and
24 : * escape characters into account according to the COPY options. The line
25 : * is copied into 'line_buf', with quotes and escape characters still
26 : * intact.
27 : *
28 : * 4. CopyReadAttributesText/CSV() function takes the input line from
29 : * 'line_buf', and splits it into fields, unescaping the data as required.
30 : * The fields are stored in 'attribute_buf', and 'raw_fields' array holds
31 : * pointers to each field.
32 : *
33 : * If encoding conversion is not required, a shortcut is taken in step 2 to
34 : * avoid copying the data unnecessarily. The 'input_buf' pointer is set to
35 : * point directly to 'raw_buf', so that CopyLoadRawBuf() loads the raw data
36 : * directly into 'input_buf'. CopyConvertBuf() then merely validates that
37 : * the data is valid in the current encoding.
38 : *
39 : * In binary mode, the pipeline is much simpler. Input is loaded into
40 : * 'raw_buf', and encoding conversion is done in the datatype-specific
41 : * receive functions, if required. 'input_buf' and 'line_buf' are not used,
42 : * but 'attribute_buf' is used as a temporary buffer to hold one attribute's
43 : * data when it's passed the receive function.
44 : *
45 : * 'raw_buf' is always 64 kB in size (RAW_BUF_SIZE). 'input_buf' is also
46 : * 64 kB (INPUT_BUF_SIZE), if encoding conversion is required. 'line_buf'
47 : * and 'attribute_buf' are expanded on demand, to hold the longest line
48 : * encountered so far.
49 : *
50 : * Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
51 : * Portions Copyright (c) 1994, Regents of the University of California
52 : *
53 : *
54 : * IDENTIFICATION
55 : * src/backend/commands/copyfromparse.c
56 : *
57 : *-------------------------------------------------------------------------
58 : */
59 : #include "postgres.h"
60 :
61 : #include <ctype.h>
62 : #include <unistd.h>
63 : #include <sys/stat.h>
64 :
65 : #include "commands/copy.h"
66 : #include "commands/copyfrom_internal.h"
67 : #include "commands/progress.h"
68 : #include "executor/executor.h"
69 : #include "libpq/libpq.h"
70 : #include "libpq/pqformat.h"
71 : #include "mb/pg_wchar.h"
72 : #include "miscadmin.h"
73 : #include "nodes/miscnodes.h"
74 : #include "pgstat.h"
75 : #include "port/pg_bswap.h"
76 : #include "utils/builtins.h"
77 : #include "utils/rel.h"
78 :
79 : #define ISOCTAL(c) (((c) >= '0') && ((c) <= '7'))
80 : #define OCTVALUE(c) ((c) - '0')
81 :
82 : /*
83 : * These macros centralize code used to process line_buf and input_buf buffers.
84 : * They are macros because they often do continue/break control and to avoid
85 : * function call overhead in tight COPY loops.
86 : *
87 : * We must use "if (1)" because the usual "do {...} while(0)" wrapper would
88 : * prevent the continue/break processing from working. We end the "if (1)"
89 : * with "else ((void) 0)" to ensure the "if" does not unintentionally match
90 : * any "else" in the calling code, and to avoid any compiler warnings about
91 : * empty statements. See http://www.cit.gu.edu.au/~anthony/info/C/C.macros.
92 : */
93 :
94 : /*
95 : * This keeps the character read at the top of the loop in the buffer
96 : * even if there is more than one read-ahead.
97 : */
98 : #define IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(extralen) \
99 : if (1) \
100 : { \
101 : if (input_buf_ptr + (extralen) >= copy_buf_len && !hit_eof) \
102 : { \
103 : input_buf_ptr = prev_raw_ptr; /* undo fetch */ \
104 : need_data = true; \
105 : continue; \
106 : } \
107 : } else ((void) 0)
108 :
109 : /* This consumes the remainder of the buffer and breaks */
110 : #define IF_NEED_REFILL_AND_EOF_BREAK(extralen) \
111 : if (1) \
112 : { \
113 : if (input_buf_ptr + (extralen) >= copy_buf_len && hit_eof) \
114 : { \
115 : if (extralen) \
116 : input_buf_ptr = copy_buf_len; /* consume the partial character */ \
117 : /* backslash just before EOF, treat as data char */ \
118 : result = true; \
119 : break; \
120 : } \
121 : } else ((void) 0)
122 :
123 : /*
124 : * Transfer any approved data to line_buf; must do this to be sure
125 : * there is some room in input_buf.
126 : */
127 : #define REFILL_LINEBUF \
128 : if (1) \
129 : { \
130 : if (input_buf_ptr > cstate->input_buf_index) \
131 : { \
132 : appendBinaryStringInfo(&cstate->line_buf, \
133 : cstate->input_buf + cstate->input_buf_index, \
134 : input_buf_ptr - cstate->input_buf_index); \
135 : cstate->input_buf_index = input_buf_ptr; \
136 : } \
137 : } else ((void) 0)
138 :
139 : /* Undo any read-ahead and jump out of the block. */
140 : #define NO_END_OF_COPY_GOTO \
141 : if (1) \
142 : { \
143 : input_buf_ptr = prev_raw_ptr + 1; \
144 : goto not_end_of_copy; \
145 : } else ((void) 0)
146 :
147 : /* NOTE: there's a copy of this in copyto.c */
148 : static const char BinarySignature[11] = "PGCOPY\n\377\r\n\0";
149 :
150 :
151 : /* non-export function prototypes */
152 : static bool CopyReadLine(CopyFromState cstate);
153 : static bool CopyReadLineText(CopyFromState cstate);
154 : static int CopyReadAttributesText(CopyFromState cstate);
155 : static int CopyReadAttributesCSV(CopyFromState cstate);
156 : static Datum CopyReadBinaryAttribute(CopyFromState cstate, FmgrInfo *flinfo,
157 : Oid typioparam, int32 typmod,
158 : bool *isnull);
159 :
160 :
161 : /* Low-level communications functions */
162 : static int CopyGetData(CopyFromState cstate, void *databuf,
163 : int minread, int maxread);
164 : static inline bool CopyGetInt32(CopyFromState cstate, int32 *val);
165 : static inline bool CopyGetInt16(CopyFromState cstate, int16 *val);
166 : static void CopyLoadInputBuf(CopyFromState cstate);
167 : static int CopyReadBinaryData(CopyFromState cstate, char *dest, int nbytes);
168 :
169 : void
170 890 : ReceiveCopyBegin(CopyFromState cstate)
171 : {
172 : StringInfoData buf;
173 890 : int natts = list_length(cstate->attnumlist);
174 890 : int16 format = (cstate->opts.binary ? 1 : 0);
175 : int i;
176 :
177 890 : pq_beginmessage(&buf, PqMsg_CopyInResponse);
178 890 : pq_sendbyte(&buf, format); /* overall format */
179 890 : pq_sendint16(&buf, natts);
180 3000 : for (i = 0; i < natts; i++)
181 2110 : pq_sendint16(&buf, format); /* per-column formats */
182 890 : pq_endmessage(&buf);
183 890 : cstate->copy_src = COPY_FRONTEND;
184 890 : cstate->fe_msgbuf = makeStringInfo();
185 : /* We *must* flush here to ensure FE knows it can send. */
186 890 : pq_flush();
187 890 : }
188 :
189 : void
190 14 : ReceiveCopyBinaryHeader(CopyFromState cstate)
191 : {
192 : char readSig[11];
193 : int32 tmp;
194 :
195 : /* Signature */
196 14 : if (CopyReadBinaryData(cstate, readSig, 11) != 11 ||
197 14 : memcmp(readSig, BinarySignature, 11) != 0)
198 0 : ereport(ERROR,
199 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
200 : errmsg("COPY file signature not recognized")));
201 : /* Flags field */
202 14 : if (!CopyGetInt32(cstate, &tmp))
203 0 : ereport(ERROR,
204 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
205 : errmsg("invalid COPY file header (missing flags)")));
206 14 : if ((tmp & (1 << 16)) != 0)
207 0 : ereport(ERROR,
208 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
209 : errmsg("invalid COPY file header (WITH OIDS)")));
210 14 : tmp &= ~(1 << 16);
211 14 : if ((tmp >> 16) != 0)
212 0 : ereport(ERROR,
213 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
214 : errmsg("unrecognized critical flags in COPY file header")));
215 : /* Header extension length */
216 14 : if (!CopyGetInt32(cstate, &tmp) ||
217 14 : tmp < 0)
218 0 : ereport(ERROR,
219 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
220 : errmsg("invalid COPY file header (missing length)")));
221 : /* Skip extension header, if present */
222 14 : while (tmp-- > 0)
223 : {
224 0 : if (CopyReadBinaryData(cstate, readSig, 1) != 1)
225 0 : ereport(ERROR,
226 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
227 : errmsg("invalid COPY file header (wrong length)")));
228 : }
229 14 : }
230 :
231 : /*
232 : * CopyGetData reads data from the source (file or frontend)
233 : *
234 : * We attempt to read at least minread, and at most maxread, bytes from
235 : * the source. The actual number of bytes read is returned; if this is
236 : * less than minread, EOF was detected.
237 : *
238 : * Note: when copying from the frontend, we expect a proper EOF mark per
239 : * protocol; if the frontend simply drops the connection, we raise error.
240 : * It seems unwise to allow the COPY IN to complete normally in that case.
241 : *
242 : * NB: no data conversion is applied here.
243 : */
244 : static int
245 430786 : CopyGetData(CopyFromState cstate, void *databuf, int minread, int maxread)
246 : {
247 430786 : int bytesread = 0;
248 :
249 430786 : switch (cstate->copy_src)
250 : {
251 970 : case COPY_FILE:
252 970 : bytesread = fread(databuf, 1, maxread, cstate->copy_file);
253 970 : if (ferror(cstate->copy_file))
254 0 : ereport(ERROR,
255 : (errcode_for_file_access(),
256 : errmsg("could not read from COPY file: %m")));
257 970 : if (bytesread == 0)
258 368 : cstate->raw_reached_eof = true;
259 970 : break;
260 401858 : case COPY_FRONTEND:
261 802924 : while (maxread > 0 && bytesread < minread && !cstate->raw_reached_eof)
262 : {
263 : int avail;
264 :
265 401778 : while (cstate->fe_msgbuf->cursor >= cstate->fe_msgbuf->len)
266 : {
267 : /* Try to receive another message */
268 : int mtype;
269 : int maxmsglen;
270 :
271 401778 : readmessage:
272 401778 : HOLD_CANCEL_INTERRUPTS();
273 401778 : pq_startmsgread();
274 401778 : mtype = pq_getbyte();
275 401778 : if (mtype == EOF)
276 0 : ereport(ERROR,
277 : (errcode(ERRCODE_CONNECTION_FAILURE),
278 : errmsg("unexpected EOF on client connection with an open transaction")));
279 : /* Validate message type and set packet size limit */
280 : switch (mtype)
281 : {
282 401066 : case PqMsg_CopyData:
283 401066 : maxmsglen = PQ_LARGE_MESSAGE_LIMIT;
284 401066 : break;
285 712 : case PqMsg_CopyDone:
286 : case PqMsg_CopyFail:
287 : case PqMsg_Flush:
288 : case PqMsg_Sync:
289 712 : maxmsglen = PQ_SMALL_MESSAGE_LIMIT;
290 712 : break;
291 0 : default:
292 0 : ereport(ERROR,
293 : (errcode(ERRCODE_PROTOCOL_VIOLATION),
294 : errmsg("unexpected message type 0x%02X during COPY from stdin",
295 : mtype)));
296 : maxmsglen = 0; /* keep compiler quiet */
297 : break;
298 : }
299 : /* Now collect the message body */
300 401778 : if (pq_getmessage(cstate->fe_msgbuf, maxmsglen))
301 0 : ereport(ERROR,
302 : (errcode(ERRCODE_CONNECTION_FAILURE),
303 : errmsg("unexpected EOF on client connection with an open transaction")));
304 401778 : RESUME_CANCEL_INTERRUPTS();
305 : /* ... and process it */
306 : switch (mtype)
307 : {
308 401066 : case PqMsg_CopyData:
309 401066 : break;
310 712 : case PqMsg_CopyDone:
311 : /* COPY IN correctly terminated by frontend */
312 712 : cstate->raw_reached_eof = true;
313 712 : return bytesread;
314 0 : case PqMsg_CopyFail:
315 0 : ereport(ERROR,
316 : (errcode(ERRCODE_QUERY_CANCELED),
317 : errmsg("COPY from stdin failed: %s",
318 : pq_getmsgstring(cstate->fe_msgbuf))));
319 : break;
320 0 : case PqMsg_Flush:
321 : case PqMsg_Sync:
322 :
323 : /*
324 : * Ignore Flush/Sync for the convenience of client
325 : * libraries (such as libpq) that may send those
326 : * without noticing that the command they just
327 : * sent was COPY.
328 : */
329 0 : goto readmessage;
330 802844 : default:
331 : Assert(false); /* NOT REACHED */
332 : }
333 : }
334 401066 : avail = cstate->fe_msgbuf->len - cstate->fe_msgbuf->cursor;
335 401066 : if (avail > maxread)
336 0 : avail = maxread;
337 401066 : pq_copymsgbytes(cstate->fe_msgbuf, databuf, avail);
338 401066 : databuf = (void *) ((char *) databuf + avail);
339 401066 : maxread -= avail;
340 401066 : bytesread += avail;
341 : }
342 401146 : break;
343 27958 : case COPY_CALLBACK:
344 27958 : bytesread = cstate->data_source_cb(databuf, minread, maxread);
345 27958 : break;
346 : }
347 :
348 430074 : return bytesread;
349 : }
350 :
351 :
352 : /*
353 : * These functions do apply some data conversion
354 : */
355 :
356 : /*
357 : * CopyGetInt32 reads an int32 that appears in network byte order
358 : *
359 : * Returns true if OK, false if EOF
360 : */
361 : static inline bool
362 186 : CopyGetInt32(CopyFromState cstate, int32 *val)
363 : {
364 : uint32 buf;
365 :
366 186 : if (CopyReadBinaryData(cstate, (char *) &buf, sizeof(buf)) != sizeof(buf))
367 : {
368 0 : *val = 0; /* suppress compiler warning */
369 0 : return false;
370 : }
371 186 : *val = (int32) pg_ntoh32(buf);
372 186 : return true;
373 : }
374 :
375 : /*
376 : * CopyGetInt16 reads an int16 that appears in network byte order
377 : */
378 : static inline bool
379 42 : CopyGetInt16(CopyFromState cstate, int16 *val)
380 : {
381 : uint16 buf;
382 :
383 42 : if (CopyReadBinaryData(cstate, (char *) &buf, sizeof(buf)) != sizeof(buf))
384 : {
385 0 : *val = 0; /* suppress compiler warning */
386 0 : return false;
387 : }
388 42 : *val = (int16) pg_ntoh16(buf);
389 42 : return true;
390 : }
391 :
392 :
393 : /*
394 : * Perform encoding conversion on data in 'raw_buf', writing the converted
395 : * data into 'input_buf'.
396 : *
397 : * On entry, there must be some data to convert in 'raw_buf'.
398 : */
399 : static void
400 860076 : CopyConvertBuf(CopyFromState cstate)
401 : {
402 : /*
403 : * If the file and server encoding are the same, no encoding conversion is
404 : * required. However, we still need to verify that the input is valid for
405 : * the encoding.
406 : */
407 860076 : if (!cstate->need_transcoding)
408 : {
409 : /*
410 : * When conversion is not required, input_buf and raw_buf are the
411 : * same. raw_buf_len is the total number of bytes in the buffer, and
412 : * input_buf_len tracks how many of those bytes have already been
413 : * verified.
414 : */
415 860076 : int preverifiedlen = cstate->input_buf_len;
416 860076 : int unverifiedlen = cstate->raw_buf_len - cstate->input_buf_len;
417 : int nverified;
418 :
419 860076 : if (unverifiedlen == 0)
420 : {
421 : /*
422 : * If no more raw data is coming, report the EOF to the caller.
423 : */
424 430810 : if (cstate->raw_reached_eof)
425 772 : cstate->input_reached_eof = true;
426 430810 : return;
427 : }
428 :
429 : /*
430 : * Verify the new data, including any residual unverified bytes from
431 : * previous round.
432 : */
433 429266 : nverified = pg_encoding_verifymbstr(cstate->file_encoding,
434 429266 : cstate->raw_buf + preverifiedlen,
435 : unverifiedlen);
436 429266 : if (nverified == 0)
437 : {
438 : /*
439 : * Could not verify anything.
440 : *
441 : * If there is no more raw input data coming, it means that there
442 : * was an incomplete multi-byte sequence at the end. Also, if
443 : * there's "enough" input left, we should be able to verify at
444 : * least one character, and a failure to do so means that we've
445 : * hit an invalid byte sequence.
446 : */
447 0 : if (cstate->raw_reached_eof || unverifiedlen >= pg_encoding_max_length(cstate->file_encoding))
448 0 : cstate->input_reached_error = true;
449 0 : return;
450 : }
451 429266 : cstate->input_buf_len += nverified;
452 : }
453 : else
454 : {
455 : /*
456 : * Encoding conversion is needed.
457 : */
458 : int nbytes;
459 : unsigned char *src;
460 : int srclen;
461 : unsigned char *dst;
462 : int dstlen;
463 : int convertedlen;
464 :
465 0 : if (RAW_BUF_BYTES(cstate) == 0)
466 : {
467 : /*
468 : * If no more raw data is coming, report the EOF to the caller.
469 : */
470 0 : if (cstate->raw_reached_eof)
471 0 : cstate->input_reached_eof = true;
472 0 : return;
473 : }
474 :
475 : /*
476 : * First, copy down any unprocessed data.
477 : */
478 0 : nbytes = INPUT_BUF_BYTES(cstate);
479 0 : if (nbytes > 0 && cstate->input_buf_index > 0)
480 0 : memmove(cstate->input_buf, cstate->input_buf + cstate->input_buf_index,
481 : nbytes);
482 0 : cstate->input_buf_index = 0;
483 0 : cstate->input_buf_len = nbytes;
484 0 : cstate->input_buf[nbytes] = '\0';
485 :
486 0 : src = (unsigned char *) cstate->raw_buf + cstate->raw_buf_index;
487 0 : srclen = cstate->raw_buf_len - cstate->raw_buf_index;
488 0 : dst = (unsigned char *) cstate->input_buf + cstate->input_buf_len;
489 0 : dstlen = INPUT_BUF_SIZE - cstate->input_buf_len + 1;
490 :
491 : /*
492 : * Do the conversion. This might stop short, if there is an invalid
493 : * byte sequence in the input. We'll convert as much as we can in
494 : * that case.
495 : *
496 : * Note: Even if we hit an invalid byte sequence, we don't report the
497 : * error until all the valid bytes have been consumed. The input
498 : * might contain an end-of-input marker (\.), and we don't want to
499 : * report an error if the invalid byte sequence is after the
500 : * end-of-input marker. We might unnecessarily convert some data
501 : * after the end-of-input marker as long as it's valid for the
502 : * encoding, but that's harmless.
503 : */
504 0 : convertedlen = pg_do_encoding_conversion_buf(cstate->conversion_proc,
505 : cstate->file_encoding,
506 : GetDatabaseEncoding(),
507 : src, srclen,
508 : dst, dstlen,
509 : true);
510 0 : if (convertedlen == 0)
511 : {
512 : /*
513 : * Could not convert anything. If there is no more raw input data
514 : * coming, it means that there was an incomplete multi-byte
515 : * sequence at the end. Also, if there is plenty of input left,
516 : * we should be able to convert at least one character, so a
517 : * failure to do so must mean that we've hit a byte sequence
518 : * that's invalid.
519 : */
520 0 : if (cstate->raw_reached_eof || srclen >= MAX_CONVERSION_INPUT_LENGTH)
521 0 : cstate->input_reached_error = true;
522 0 : return;
523 : }
524 0 : cstate->raw_buf_index += convertedlen;
525 0 : cstate->input_buf_len += strlen((char *) dst);
526 : }
527 : }
528 :
529 : /*
530 : * Report an encoding or conversion error.
531 : */
532 : static void
533 0 : CopyConversionError(CopyFromState cstate)
534 : {
535 : Assert(cstate->raw_buf_len > 0);
536 : Assert(cstate->input_reached_error);
537 :
538 0 : if (!cstate->need_transcoding)
539 : {
540 : /*
541 : * Everything up to input_buf_len was successfully verified, and
542 : * input_buf_len points to the invalid or incomplete character.
543 : */
544 0 : report_invalid_encoding(cstate->file_encoding,
545 0 : cstate->raw_buf + cstate->input_buf_len,
546 0 : cstate->raw_buf_len - cstate->input_buf_len);
547 : }
548 : else
549 : {
550 : /*
551 : * raw_buf_index points to the invalid or untranslatable character. We
552 : * let the conversion routine report the error, because it can provide
553 : * a more specific error message than we could here. An earlier call
554 : * to the conversion routine in CopyConvertBuf() detected that there
555 : * is an error, now we call the conversion routine again with
556 : * noError=false, to have it throw the error.
557 : */
558 : unsigned char *src;
559 : int srclen;
560 : unsigned char *dst;
561 : int dstlen;
562 :
563 0 : src = (unsigned char *) cstate->raw_buf + cstate->raw_buf_index;
564 0 : srclen = cstate->raw_buf_len - cstate->raw_buf_index;
565 0 : dst = (unsigned char *) cstate->input_buf + cstate->input_buf_len;
566 0 : dstlen = INPUT_BUF_SIZE - cstate->input_buf_len + 1;
567 :
568 0 : (void) pg_do_encoding_conversion_buf(cstate->conversion_proc,
569 : cstate->file_encoding,
570 : GetDatabaseEncoding(),
571 : src, srclen,
572 : dst, dstlen,
573 : false);
574 :
575 : /*
576 : * The conversion routine should have reported an error, so this
577 : * should not be reached.
578 : */
579 0 : elog(ERROR, "encoding conversion failed without error");
580 : }
581 : }
582 :
583 : /*
584 : * Load more data from data source to raw_buf.
585 : *
586 : * If RAW_BUF_BYTES(cstate) > 0, the unprocessed bytes are moved to the
587 : * beginning of the buffer, and we load new data after that.
588 : */
589 : static void
590 430074 : CopyLoadRawBuf(CopyFromState cstate)
591 : {
592 : int nbytes;
593 : int inbytes;
594 :
595 : /*
596 : * In text mode, if encoding conversion is not required, raw_buf and
597 : * input_buf point to the same buffer. Their len/index better agree, too.
598 : */
599 430074 : if (cstate->raw_buf == cstate->input_buf)
600 : {
601 : Assert(!cstate->need_transcoding);
602 : Assert(cstate->raw_buf_index == cstate->input_buf_index);
603 : Assert(cstate->input_buf_len <= cstate->raw_buf_len);
604 : }
605 :
606 : /*
607 : * Copy down the unprocessed data if any.
608 : */
609 430074 : nbytes = RAW_BUF_BYTES(cstate);
610 430074 : if (nbytes > 0 && cstate->raw_buf_index > 0)
611 0 : memmove(cstate->raw_buf, cstate->raw_buf + cstate->raw_buf_index,
612 : nbytes);
613 430074 : cstate->raw_buf_len -= cstate->raw_buf_index;
614 430074 : cstate->raw_buf_index = 0;
615 :
616 : /*
617 : * If raw_buf and input_buf are in fact the same buffer, adjust the
618 : * input_buf variables, too.
619 : */
620 430074 : if (cstate->raw_buf == cstate->input_buf)
621 : {
622 430038 : cstate->input_buf_len -= cstate->input_buf_index;
623 430038 : cstate->input_buf_index = 0;
624 : }
625 :
626 : /* Load more data */
627 430074 : inbytes = CopyGetData(cstate, cstate->raw_buf + cstate->raw_buf_len,
628 430074 : 1, RAW_BUF_SIZE - cstate->raw_buf_len);
629 430074 : nbytes += inbytes;
630 430074 : cstate->raw_buf[nbytes] = '\0';
631 430074 : cstate->raw_buf_len = nbytes;
632 :
633 430074 : cstate->bytes_processed += inbytes;
634 430074 : pgstat_progress_update_param(PROGRESS_COPY_BYTES_PROCESSED, cstate->bytes_processed);
635 :
636 430074 : if (inbytes == 0)
637 784 : cstate->raw_reached_eof = true;
638 430074 : }
639 :
640 : /*
641 : * CopyLoadInputBuf loads some more data into input_buf
642 : *
643 : * On return, at least one more input character is loaded into
644 : * input_buf, or input_reached_eof is set.
645 : *
646 : * If INPUT_BUF_BYTES(cstate) > 0, the unprocessed bytes are moved to the start
647 : * of the buffer and then we load more data after that.
648 : */
649 : static void
650 430038 : CopyLoadInputBuf(CopyFromState cstate)
651 : {
652 430038 : int nbytes = INPUT_BUF_BYTES(cstate);
653 :
654 : /*
655 : * The caller has updated input_buf_index to indicate how much of the
656 : * input has been consumed and isn't needed anymore. If input_buf is the
657 : * same physical area as raw_buf, update raw_buf_index accordingly.
658 : */
659 430038 : if (cstate->raw_buf == cstate->input_buf)
660 : {
661 : Assert(!cstate->need_transcoding);
662 : Assert(cstate->input_buf_index >= cstate->raw_buf_index);
663 430038 : cstate->raw_buf_index = cstate->input_buf_index;
664 : }
665 :
666 : for (;;)
667 : {
668 : /* If we now have some unconverted data, try to convert it */
669 860076 : CopyConvertBuf(cstate);
670 :
671 : /* If we now have some more input bytes ready, return them */
672 860076 : if (INPUT_BUF_BYTES(cstate) > nbytes)
673 429266 : return;
674 :
675 : /*
676 : * If we reached an invalid byte sequence, or we're at an incomplete
677 : * multi-byte character but there is no more raw input data, report
678 : * conversion error.
679 : */
680 430810 : if (cstate->input_reached_error)
681 0 : CopyConversionError(cstate);
682 :
683 : /* no more input, and everything has been converted */
684 430810 : if (cstate->input_reached_eof)
685 772 : break;
686 :
687 : /* Try to load more raw data */
688 : Assert(!cstate->raw_reached_eof);
689 430038 : CopyLoadRawBuf(cstate);
690 : }
691 : }
692 :
693 : /*
694 : * CopyReadBinaryData
695 : *
696 : * Reads up to 'nbytes' bytes from cstate->copy_file via cstate->raw_buf
697 : * and writes them to 'dest'. Returns the number of bytes read (which
698 : * would be less than 'nbytes' only if we reach EOF).
699 : */
700 : static int
701 382 : CopyReadBinaryData(CopyFromState cstate, char *dest, int nbytes)
702 : {
703 382 : int copied_bytes = 0;
704 :
705 382 : if (RAW_BUF_BYTES(cstate) >= nbytes)
706 : {
707 : /* Enough bytes are present in the buffer. */
708 346 : memcpy(dest, cstate->raw_buf + cstate->raw_buf_index, nbytes);
709 346 : cstate->raw_buf_index += nbytes;
710 346 : copied_bytes = nbytes;
711 : }
712 : else
713 : {
714 : /*
715 : * Not enough bytes in the buffer, so must read from the file. Need
716 : * to loop since 'nbytes' could be larger than the buffer size.
717 : */
718 : do
719 : {
720 : int copy_bytes;
721 :
722 : /* Load more data if buffer is empty. */
723 36 : if (RAW_BUF_BYTES(cstate) == 0)
724 : {
725 36 : CopyLoadRawBuf(cstate);
726 36 : if (cstate->raw_reached_eof)
727 12 : break; /* EOF */
728 : }
729 :
730 : /* Transfer some bytes. */
731 24 : copy_bytes = Min(nbytes - copied_bytes, RAW_BUF_BYTES(cstate));
732 24 : memcpy(dest, cstate->raw_buf + cstate->raw_buf_index, copy_bytes);
733 24 : cstate->raw_buf_index += copy_bytes;
734 24 : dest += copy_bytes;
735 24 : copied_bytes += copy_bytes;
736 24 : } while (copied_bytes < nbytes);
737 : }
738 :
739 382 : return copied_bytes;
740 : }
741 :
742 : /*
743 : * Read raw fields in the next line for COPY FROM in text or csv mode.
744 : * Return false if no more lines.
745 : *
746 : * An internal temporary buffer is returned via 'fields'. It is valid until
747 : * the next call of the function. Since the function returns all raw fields
748 : * in the input file, 'nfields' could be different from the number of columns
749 : * in the relation.
750 : *
751 : * NOTE: force_not_null option are not applied to the returned fields.
752 : */
753 : bool
754 1250174 : NextCopyFromRawFields(CopyFromState cstate, char ***fields, int *nfields)
755 : {
756 : int fldct;
757 : bool done;
758 :
759 : /* only available for text or csv input */
760 : Assert(!cstate->opts.binary);
761 :
762 : /* on input check that the header line is correct if needed */
763 1250174 : if (cstate->cur_lineno == 0 && cstate->opts.header_line)
764 : {
765 : ListCell *cur;
766 : TupleDesc tupDesc;
767 :
768 110 : tupDesc = RelationGetDescr(cstate->rel);
769 :
770 110 : cstate->cur_lineno++;
771 110 : done = CopyReadLine(cstate);
772 :
773 110 : if (cstate->opts.header_line == COPY_HEADER_MATCH)
774 : {
775 : int fldnum;
776 :
777 76 : if (cstate->opts.csv_mode)
778 10 : fldct = CopyReadAttributesCSV(cstate);
779 : else
780 66 : fldct = CopyReadAttributesText(cstate);
781 :
782 76 : if (fldct != list_length(cstate->attnumlist))
783 24 : ereport(ERROR,
784 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
785 : errmsg("wrong number of fields in header line: got %d, expected %d",
786 : fldct, list_length(cstate->attnumlist))));
787 :
788 52 : fldnum = 0;
789 158 : foreach(cur, cstate->attnumlist)
790 : {
791 126 : int attnum = lfirst_int(cur);
792 : char *colName;
793 126 : Form_pg_attribute attr = TupleDescAttr(tupDesc, attnum - 1);
794 :
795 : Assert(fldnum < cstate->max_fields);
796 :
797 126 : colName = cstate->raw_fields[fldnum++];
798 126 : if (colName == NULL)
799 6 : ereport(ERROR,
800 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
801 : errmsg("column name mismatch in header line field %d: got null value (\"%s\"), expected \"%s\"",
802 : fldnum, cstate->opts.null_print, NameStr(attr->attname))));
803 :
804 120 : if (namestrcmp(&attr->attname, colName) != 0)
805 : {
806 14 : ereport(ERROR,
807 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
808 : errmsg("column name mismatch in header line field %d: got \"%s\", expected \"%s\"",
809 : fldnum, colName, NameStr(attr->attname))));
810 : }
811 : }
812 : }
813 :
814 66 : if (done)
815 0 : return false;
816 : }
817 :
818 1250130 : cstate->cur_lineno++;
819 :
820 : /* Actually read the line into memory here */
821 1250130 : done = CopyReadLine(cstate);
822 :
823 : /*
824 : * EOF at start of line means we're done. If we see EOF after some
825 : * characters, we act as though it was newline followed by EOF, ie,
826 : * process the line and then exit loop on next iteration.
827 : */
828 1250130 : if (done && cstate->line_buf.len == 0)
829 1404 : return false;
830 :
831 : /* Parse the line into de-escaped field values */
832 1248726 : if (cstate->opts.csv_mode)
833 394 : fldct = CopyReadAttributesCSV(cstate);
834 : else
835 1248332 : fldct = CopyReadAttributesText(cstate);
836 :
837 1248714 : *fields = cstate->raw_fields;
838 1248714 : *nfields = fldct;
839 1248714 : return true;
840 : }
841 :
842 : /*
843 : * Read next tuple from file for COPY FROM. Return false if no more tuples.
844 : *
845 : * 'econtext' is used to evaluate default expression for each column that is
846 : * either not read from the file or is using the DEFAULT option of COPY FROM.
847 : * It can be NULL when no default values are used, i.e. when all columns are
848 : * read from the file, and DEFAULT option is unset.
849 : *
850 : * 'values' and 'nulls' arrays must be the same length as columns of the
851 : * relation passed to BeginCopyFrom. This function fills the arrays.
852 : */
853 : bool
854 1250216 : NextCopyFrom(CopyFromState cstate, ExprContext *econtext,
855 : Datum *values, bool *nulls)
856 : {
857 : TupleDesc tupDesc;
858 : AttrNumber num_phys_attrs,
859 : attr_count,
860 1250216 : num_defaults = cstate->num_defaults;
861 1250216 : FmgrInfo *in_functions = cstate->in_functions;
862 1250216 : Oid *typioparams = cstate->typioparams;
863 : int i;
864 1250216 : int *defmap = cstate->defmap;
865 1250216 : ExprState **defexprs = cstate->defexprs;
866 :
867 1250216 : tupDesc = RelationGetDescr(cstate->rel);
868 1250216 : num_phys_attrs = tupDesc->natts;
869 1250216 : attr_count = list_length(cstate->attnumlist);
870 :
871 : /* Initialize all values for row to NULL */
872 5801616 : MemSet(values, 0, num_phys_attrs * sizeof(Datum));
873 1250216 : MemSet(nulls, true, num_phys_attrs * sizeof(bool));
874 1394324 : MemSet(cstate->defaults, false, num_phys_attrs * sizeof(bool));
875 :
876 1250216 : if (!cstate->opts.binary)
877 : {
878 : char **field_strings;
879 : ListCell *cur;
880 : int fldct;
881 : int fieldno;
882 : char *string;
883 :
884 : /* read raw fields in the next line */
885 1250174 : if (!NextCopyFromRawFields(cstate, &field_strings, &fldct))
886 1446 : return false;
887 :
888 : /* check for overflowing fields */
889 1248714 : if (attr_count > 0 && fldct > attr_count)
890 18 : ereport(ERROR,
891 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
892 : errmsg("extra data after last expected column")));
893 :
894 1248696 : fieldno = 0;
895 :
896 : /* Loop to read the user attributes on the line. */
897 5677052 : foreach(cur, cstate->attnumlist)
898 : {
899 4428454 : int attnum = lfirst_int(cur);
900 4428454 : int m = attnum - 1;
901 4428454 : Form_pg_attribute att = TupleDescAttr(tupDesc, m);
902 :
903 4428454 : if (fieldno >= fldct)
904 18 : ereport(ERROR,
905 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
906 : errmsg("missing data for column \"%s\"",
907 : NameStr(att->attname))));
908 4428436 : string = field_strings[fieldno++];
909 :
910 4428436 : if (cstate->convert_select_flags &&
911 20 : !cstate->convert_select_flags[m])
912 : {
913 : /* ignore input field, leaving column as NULL */
914 10 : continue;
915 : }
916 :
917 4428426 : if (cstate->opts.csv_mode)
918 : {
919 852 : if (string == NULL &&
920 44 : cstate->opts.force_notnull_flags[m])
921 : {
922 : /*
923 : * FORCE_NOT_NULL option is set and column is NULL -
924 : * convert it to the NULL string.
925 : */
926 28 : string = cstate->opts.null_print;
927 : }
928 824 : else if (string != NULL && cstate->opts.force_null_flags[m]
929 50 : && strcmp(string, cstate->opts.null_print) == 0)
930 : {
931 : /*
932 : * FORCE_NULL option is set and column matches the NULL
933 : * string. It must have been quoted, or otherwise the
934 : * string would already have been set to NULL. Convert it
935 : * to NULL as specified.
936 : */
937 26 : string = NULL;
938 : }
939 : }
940 :
941 4428426 : cstate->cur_attname = NameStr(att->attname);
942 4428426 : cstate->cur_attval = string;
943 :
944 4428426 : if (string != NULL)
945 4423588 : nulls[m] = false;
946 :
947 4428426 : if (cstate->defaults[m])
948 : {
949 : /*
950 : * The caller must supply econtext and have switched into the
951 : * per-tuple memory context in it.
952 : */
953 : Assert(econtext != NULL);
954 : Assert(CurrentMemoryContext == econtext->ecxt_per_tuple_memory);
955 :
956 60 : values[m] = ExecEvalExpr(defexprs[m], econtext, &nulls[m]);
957 : }
958 :
959 : /*
960 : * If ON_ERROR is specified with IGNORE, skip rows with soft
961 : * errors
962 : */
963 4428328 : else if (!InputFunctionCallSafe(&in_functions[m],
964 : string,
965 4428366 : typioparams[m],
966 : att->atttypmod,
967 4428366 : (Node *) cstate->escontext,
968 4428366 : &values[m]))
969 : {
970 : Assert(cstate->opts.on_error != COPY_ON_ERROR_STOP);
971 :
972 42 : cstate->num_errors++;
973 :
974 42 : if (cstate->opts.log_verbosity == COPY_LOG_VERBOSITY_VERBOSE)
975 : {
976 : /*
977 : * Since we emit line number and column info in the below
978 : * notice message, we suppress error context information
979 : * other than the relation name.
980 : */
981 : Assert(!cstate->relname_only);
982 42 : cstate->relname_only = true;
983 :
984 42 : if (cstate->cur_attval)
985 : {
986 : char *attval;
987 :
988 36 : attval = CopyLimitPrintoutLength(cstate->cur_attval);
989 36 : ereport(NOTICE,
990 : errmsg("skipping row due to data type incompatibility at line %llu for column \"%s\": \"%s\"",
991 : (unsigned long long) cstate->cur_lineno,
992 : cstate->cur_attname,
993 : attval));
994 36 : pfree(attval);
995 : }
996 : else
997 6 : ereport(NOTICE,
998 : errmsg("skipping row due to data type incompatibility at line %llu for column \"%s\": null input",
999 : (unsigned long long) cstate->cur_lineno,
1000 : cstate->cur_attname));
1001 :
1002 : /* reset relname_only */
1003 42 : cstate->relname_only = false;
1004 : }
1005 :
1006 42 : return true;
1007 : }
1008 :
1009 4428346 : cstate->cur_attname = NULL;
1010 4428346 : cstate->cur_attval = NULL;
1011 : }
1012 :
1013 : Assert(fieldno == attr_count);
1014 : }
1015 : else
1016 : {
1017 : /* binary */
1018 : int16 fld_count;
1019 : ListCell *cur;
1020 :
1021 42 : cstate->cur_lineno++;
1022 :
1023 42 : if (!CopyGetInt16(cstate, &fld_count))
1024 : {
1025 : /* EOF detected (end of file, or protocol-level EOF) */
1026 12 : return false;
1027 : }
1028 :
1029 42 : if (fld_count == -1)
1030 : {
1031 : /*
1032 : * Received EOF marker. Wait for the protocol-level EOF, and
1033 : * complain if it doesn't come immediately. In COPY FROM STDIN,
1034 : * this ensures that we correctly handle CopyFail, if client
1035 : * chooses to send that now. When copying from file, we could
1036 : * ignore the rest of the file like in text mode, but we choose to
1037 : * be consistent with the COPY FROM STDIN case.
1038 : */
1039 : char dummy;
1040 :
1041 12 : if (CopyReadBinaryData(cstate, &dummy, 1) > 0)
1042 0 : ereport(ERROR,
1043 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1044 : errmsg("received copy data after EOF marker")));
1045 12 : return false;
1046 : }
1047 :
1048 30 : if (fld_count != attr_count)
1049 0 : ereport(ERROR,
1050 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1051 : errmsg("row field count is %d, expected %d",
1052 : (int) fld_count, attr_count)));
1053 :
1054 186 : foreach(cur, cstate->attnumlist)
1055 : {
1056 158 : int attnum = lfirst_int(cur);
1057 158 : int m = attnum - 1;
1058 158 : Form_pg_attribute att = TupleDescAttr(tupDesc, m);
1059 :
1060 158 : cstate->cur_attname = NameStr(att->attname);
1061 314 : values[m] = CopyReadBinaryAttribute(cstate,
1062 158 : &in_functions[m],
1063 158 : typioparams[m],
1064 : att->atttypmod,
1065 : &nulls[m]);
1066 156 : cstate->cur_attname = NULL;
1067 : }
1068 : }
1069 :
1070 : /*
1071 : * Now compute and insert any defaults available for the columns not
1072 : * provided by the input data. Anything not processed here or above will
1073 : * remain NULL.
1074 : */
1075 1309156 : for (i = 0; i < num_defaults; i++)
1076 : {
1077 : /*
1078 : * The caller must supply econtext and have switched into the
1079 : * per-tuple memory context in it.
1080 : */
1081 : Assert(econtext != NULL);
1082 : Assert(CurrentMemoryContext == econtext->ecxt_per_tuple_memory);
1083 :
1084 60530 : values[defmap[i]] = ExecEvalExpr(defexprs[defmap[i]], econtext,
1085 60530 : &nulls[defmap[i]]);
1086 : }
1087 :
1088 1248626 : return true;
1089 : }
1090 :
1091 : /*
1092 : * Read the next input line and stash it in line_buf.
1093 : *
1094 : * Result is true if read was terminated by EOF, false if terminated
1095 : * by newline. The terminating newline or EOF marker is not included
1096 : * in the final value of line_buf.
1097 : */
1098 : static bool
1099 1250240 : CopyReadLine(CopyFromState cstate)
1100 : {
1101 : bool result;
1102 :
1103 1250240 : resetStringInfo(&cstate->line_buf);
1104 1250240 : cstate->line_buf_valid = false;
1105 :
1106 : /* Parse data and transfer into line_buf */
1107 1250240 : result = CopyReadLineText(cstate);
1108 :
1109 1250240 : if (result)
1110 : {
1111 : /*
1112 : * Reached EOF. In protocol version 3, we should ignore anything
1113 : * after \. up to the protocol end of copy data. (XXX maybe better
1114 : * not to treat \. as special?)
1115 : */
1116 1404 : if (cstate->copy_src == COPY_FRONTEND)
1117 : {
1118 : int inbytes;
1119 :
1120 : do
1121 : {
1122 712 : inbytes = CopyGetData(cstate, cstate->input_buf,
1123 : 1, INPUT_BUF_SIZE);
1124 712 : } while (inbytes > 0);
1125 712 : cstate->input_buf_index = 0;
1126 712 : cstate->input_buf_len = 0;
1127 712 : cstate->raw_buf_index = 0;
1128 712 : cstate->raw_buf_len = 0;
1129 : }
1130 : }
1131 : else
1132 : {
1133 : /*
1134 : * If we didn't hit EOF, then we must have transferred the EOL marker
1135 : * to line_buf along with the data. Get rid of it.
1136 : */
1137 1248836 : switch (cstate->eol_type)
1138 : {
1139 1248836 : case EOL_NL:
1140 : Assert(cstate->line_buf.len >= 1);
1141 : Assert(cstate->line_buf.data[cstate->line_buf.len - 1] == '\n');
1142 1248836 : cstate->line_buf.len--;
1143 1248836 : cstate->line_buf.data[cstate->line_buf.len] = '\0';
1144 1248836 : break;
1145 0 : case EOL_CR:
1146 : Assert(cstate->line_buf.len >= 1);
1147 : Assert(cstate->line_buf.data[cstate->line_buf.len - 1] == '\r');
1148 0 : cstate->line_buf.len--;
1149 0 : cstate->line_buf.data[cstate->line_buf.len] = '\0';
1150 0 : break;
1151 0 : case EOL_CRNL:
1152 : Assert(cstate->line_buf.len >= 2);
1153 : Assert(cstate->line_buf.data[cstate->line_buf.len - 2] == '\r');
1154 : Assert(cstate->line_buf.data[cstate->line_buf.len - 1] == '\n');
1155 0 : cstate->line_buf.len -= 2;
1156 0 : cstate->line_buf.data[cstate->line_buf.len] = '\0';
1157 0 : break;
1158 0 : case EOL_UNKNOWN:
1159 : /* shouldn't get here */
1160 : Assert(false);
1161 0 : break;
1162 : }
1163 1250240 : }
1164 :
1165 : /* Now it's safe to use the buffer in error messages */
1166 1250240 : cstate->line_buf_valid = true;
1167 :
1168 1250240 : return result;
1169 : }
1170 :
1171 : /*
1172 : * CopyReadLineText - inner loop of CopyReadLine for text mode
1173 : */
1174 : static bool
1175 1250240 : CopyReadLineText(CopyFromState cstate)
1176 : {
1177 : char *copy_input_buf;
1178 : int input_buf_ptr;
1179 : int copy_buf_len;
1180 1250240 : bool need_data = false;
1181 1250240 : bool hit_eof = false;
1182 1250240 : bool result = false;
1183 :
1184 : /* CSV variables */
1185 1250240 : bool first_char_in_line = true;
1186 1250240 : bool in_quote = false,
1187 1250240 : last_was_esc = false;
1188 1250240 : char quotec = '\0';
1189 1250240 : char escapec = '\0';
1190 :
1191 1250240 : if (cstate->opts.csv_mode)
1192 : {
1193 602 : quotec = cstate->opts.quote[0];
1194 602 : escapec = cstate->opts.escape[0];
1195 : /* ignore special escape processing if it's the same as quotec */
1196 602 : if (quotec == escapec)
1197 466 : escapec = '\0';
1198 : }
1199 :
1200 : /*
1201 : * The objective of this loop is to transfer the entire next input line
1202 : * into line_buf. Hence, we only care for detecting newlines (\r and/or
1203 : * \n) and the end-of-copy marker (\.).
1204 : *
1205 : * In CSV mode, \r and \n inside a quoted field are just part of the data
1206 : * value and are put in line_buf. We keep just enough state to know if we
1207 : * are currently in a quoted field or not.
1208 : *
1209 : * The input has already been converted to the database encoding. All
1210 : * supported server encodings have the property that all bytes in a
1211 : * multi-byte sequence have the high bit set, so a multibyte character
1212 : * cannot contain any newline or escape characters embedded in the
1213 : * multibyte sequence. Therefore, we can process the input byte-by-byte,
1214 : * regardless of the encoding.
1215 : *
1216 : * For speed, we try to move data from input_buf to line_buf in chunks
1217 : * rather than one character at a time. input_buf_ptr points to the next
1218 : * character to examine; any characters from input_buf_index to
1219 : * input_buf_ptr have been determined to be part of the line, but not yet
1220 : * transferred to line_buf.
1221 : *
1222 : * For a little extra speed within the loop, we copy input_buf and
1223 : * input_buf_len into local variables.
1224 : */
1225 1250240 : copy_input_buf = cstate->input_buf;
1226 1250240 : input_buf_ptr = cstate->input_buf_index;
1227 1250240 : copy_buf_len = cstate->input_buf_len;
1228 :
1229 : for (;;)
1230 24753402 : {
1231 : int prev_raw_ptr;
1232 : char c;
1233 :
1234 : /*
1235 : * Load more data if needed.
1236 : *
1237 : * TODO: We could just force four bytes of read-ahead and avoid the
1238 : * many calls to IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(). That was
1239 : * unsafe with the old v2 COPY protocol, but we don't support that
1240 : * anymore.
1241 : */
1242 26003642 : if (input_buf_ptr >= copy_buf_len || need_data)
1243 : {
1244 430038 : REFILL_LINEBUF;
1245 :
1246 430038 : CopyLoadInputBuf(cstate);
1247 : /* update our local variables */
1248 430038 : hit_eof = cstate->input_reached_eof;
1249 430038 : input_buf_ptr = cstate->input_buf_index;
1250 430038 : copy_buf_len = cstate->input_buf_len;
1251 :
1252 : /*
1253 : * If we are completely out of data, break out of the loop,
1254 : * reporting EOF.
1255 : */
1256 430038 : if (INPUT_BUF_BYTES(cstate) <= 0)
1257 : {
1258 772 : result = true;
1259 772 : break;
1260 : }
1261 429266 : need_data = false;
1262 : }
1263 :
1264 : /* OK to fetch a character */
1265 26002870 : prev_raw_ptr = input_buf_ptr;
1266 26002870 : c = copy_input_buf[input_buf_ptr++];
1267 :
1268 26002870 : if (cstate->opts.csv_mode)
1269 : {
1270 : /*
1271 : * If character is '\\' or '\r', we may need to look ahead below.
1272 : * Force fetch of the next character if we don't already have it.
1273 : * We need to do this before changing CSV state, in case one of
1274 : * these characters is also the quote or escape character.
1275 : */
1276 4858 : if (c == '\\' || c == '\r')
1277 : {
1278 318 : IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(0);
1279 : }
1280 :
1281 : /*
1282 : * Dealing with quotes and escapes here is mildly tricky. If the
1283 : * quote char is also the escape char, there's no problem - we
1284 : * just use the char as a toggle. If they are different, we need
1285 : * to ensure that we only take account of an escape inside a
1286 : * quoted field and immediately preceding a quote char, and not
1287 : * the second in an escape-escape sequence.
1288 : */
1289 4858 : if (in_quote && c == escapec)
1290 48 : last_was_esc = !last_was_esc;
1291 4858 : if (c == quotec && !last_was_esc)
1292 408 : in_quote = !in_quote;
1293 4858 : if (c != escapec)
1294 4804 : last_was_esc = false;
1295 :
1296 : /*
1297 : * Updating the line count for embedded CR and/or LF chars is
1298 : * necessarily a little fragile - this test is probably about the
1299 : * best we can do. (XXX it's arguable whether we should do this
1300 : * at all --- is cur_lineno a physical or logical count?)
1301 : */
1302 4858 : if (in_quote && c == (cstate->eol_type == EOL_NL ? '\n' : '\r'))
1303 36 : cstate->cur_lineno++;
1304 : }
1305 :
1306 : /* Process \r */
1307 26002870 : if (c == '\r' && (!cstate->opts.csv_mode || !in_quote))
1308 : {
1309 : /* Check for \r\n on first line, _and_ handle \r\n. */
1310 0 : if (cstate->eol_type == EOL_UNKNOWN ||
1311 0 : cstate->eol_type == EOL_CRNL)
1312 : {
1313 : /*
1314 : * If need more data, go back to loop top to load it.
1315 : *
1316 : * Note that if we are at EOF, c will wind up as '\0' because
1317 : * of the guaranteed pad of input_buf.
1318 : */
1319 0 : IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(0);
1320 :
1321 : /* get next char */
1322 0 : c = copy_input_buf[input_buf_ptr];
1323 :
1324 0 : if (c == '\n')
1325 : {
1326 0 : input_buf_ptr++; /* eat newline */
1327 0 : cstate->eol_type = EOL_CRNL; /* in case not set yet */
1328 : }
1329 : else
1330 : {
1331 : /* found \r, but no \n */
1332 0 : if (cstate->eol_type == EOL_CRNL)
1333 0 : ereport(ERROR,
1334 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1335 : !cstate->opts.csv_mode ?
1336 : errmsg("literal carriage return found in data") :
1337 : errmsg("unquoted carriage return found in data"),
1338 : !cstate->opts.csv_mode ?
1339 : errhint("Use \"\\r\" to represent carriage return.") :
1340 : errhint("Use quoted CSV field to represent carriage return.")));
1341 :
1342 : /*
1343 : * if we got here, it is the first line and we didn't find
1344 : * \n, so don't consume the peeked character
1345 : */
1346 0 : cstate->eol_type = EOL_CR;
1347 : }
1348 : }
1349 0 : else if (cstate->eol_type == EOL_NL)
1350 0 : ereport(ERROR,
1351 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1352 : !cstate->opts.csv_mode ?
1353 : errmsg("literal carriage return found in data") :
1354 : errmsg("unquoted carriage return found in data"),
1355 : !cstate->opts.csv_mode ?
1356 : errhint("Use \"\\r\" to represent carriage return.") :
1357 : errhint("Use quoted CSV field to represent carriage return.")));
1358 : /* If reach here, we have found the line terminator */
1359 0 : break;
1360 : }
1361 :
1362 : /* Process \n */
1363 26002870 : if (c == '\n' && (!cstate->opts.csv_mode || !in_quote))
1364 : {
1365 1248836 : if (cstate->eol_type == EOL_CR || cstate->eol_type == EOL_CRNL)
1366 0 : ereport(ERROR,
1367 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1368 : !cstate->opts.csv_mode ?
1369 : errmsg("literal newline found in data") :
1370 : errmsg("unquoted newline found in data"),
1371 : !cstate->opts.csv_mode ?
1372 : errhint("Use \"\\n\" to represent newline.") :
1373 : errhint("Use quoted CSV field to represent newline.")));
1374 1248836 : cstate->eol_type = EOL_NL; /* in case not set yet */
1375 : /* If reach here, we have found the line terminator */
1376 1248836 : break;
1377 : }
1378 :
1379 : /*
1380 : * In CSV mode, we only recognize \. alone on a line. This is because
1381 : * \. is a valid CSV data value.
1382 : */
1383 24754034 : if (c == '\\' && (!cstate->opts.csv_mode || first_char_in_line))
1384 : {
1385 : char c2;
1386 :
1387 8632 : IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(0);
1388 8632 : IF_NEED_REFILL_AND_EOF_BREAK(0);
1389 :
1390 : /* -----
1391 : * get next character
1392 : * Note: we do not change c so if it isn't \., we can fall
1393 : * through and continue processing.
1394 : * -----
1395 : */
1396 8632 : c2 = copy_input_buf[input_buf_ptr];
1397 :
1398 8632 : if (c2 == '.')
1399 : {
1400 638 : input_buf_ptr++; /* consume the '.' */
1401 :
1402 : /*
1403 : * Note: if we loop back for more data here, it does not
1404 : * matter that the CSV state change checks are re-executed; we
1405 : * will come back here with no important state changed.
1406 : */
1407 638 : if (cstate->eol_type == EOL_CRNL)
1408 : {
1409 : /* Get the next character */
1410 0 : IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(0);
1411 : /* if hit_eof, c2 will become '\0' */
1412 0 : c2 = copy_input_buf[input_buf_ptr++];
1413 :
1414 0 : if (c2 == '\n')
1415 : {
1416 0 : if (!cstate->opts.csv_mode)
1417 0 : ereport(ERROR,
1418 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1419 : errmsg("end-of-copy marker does not match previous newline style")));
1420 : else
1421 0 : NO_END_OF_COPY_GOTO;
1422 : }
1423 0 : else if (c2 != '\r')
1424 : {
1425 0 : if (!cstate->opts.csv_mode)
1426 0 : ereport(ERROR,
1427 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1428 : errmsg("end-of-copy marker corrupt")));
1429 : else
1430 0 : NO_END_OF_COPY_GOTO;
1431 : }
1432 : }
1433 :
1434 : /* Get the next character */
1435 638 : IF_NEED_REFILL_AND_NOT_EOF_CONTINUE(0);
1436 : /* if hit_eof, c2 will become '\0' */
1437 638 : c2 = copy_input_buf[input_buf_ptr++];
1438 :
1439 638 : if (c2 != '\r' && c2 != '\n')
1440 : {
1441 6 : if (!cstate->opts.csv_mode)
1442 0 : ereport(ERROR,
1443 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1444 : errmsg("end-of-copy marker corrupt")));
1445 : else
1446 6 : NO_END_OF_COPY_GOTO;
1447 : }
1448 :
1449 632 : if ((cstate->eol_type == EOL_NL && c2 != '\n') ||
1450 632 : (cstate->eol_type == EOL_CRNL && c2 != '\n') ||
1451 632 : (cstate->eol_type == EOL_CR && c2 != '\r'))
1452 : {
1453 0 : ereport(ERROR,
1454 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1455 : errmsg("end-of-copy marker does not match previous newline style")));
1456 : }
1457 :
1458 : /*
1459 : * Transfer only the data before the \. into line_buf, then
1460 : * discard the data and the \. sequence.
1461 : */
1462 632 : if (prev_raw_ptr > cstate->input_buf_index)
1463 0 : appendBinaryStringInfo(&cstate->line_buf,
1464 0 : cstate->input_buf + cstate->input_buf_index,
1465 0 : prev_raw_ptr - cstate->input_buf_index);
1466 632 : cstate->input_buf_index = input_buf_ptr;
1467 632 : result = true; /* report EOF */
1468 632 : break;
1469 : }
1470 7994 : else if (!cstate->opts.csv_mode)
1471 : {
1472 : /*
1473 : * If we are here, it means we found a backslash followed by
1474 : * something other than a period. In non-CSV mode, anything
1475 : * after a backslash is special, so we skip over that second
1476 : * character too. If we didn't do that \\. would be
1477 : * considered an eof-of copy, while in non-CSV mode it is a
1478 : * literal backslash followed by a period. In CSV mode,
1479 : * backslashes are not special, so we want to process the
1480 : * character after the backslash just like a normal character,
1481 : * so we don't increment in those cases.
1482 : */
1483 7988 : input_buf_ptr++;
1484 : }
1485 : }
1486 :
1487 : /*
1488 : * This label is for CSV cases where \. appears at the start of a
1489 : * line, but there is more text after it, meaning it was a data value.
1490 : * We are more strict for \. in CSV mode because \. could be a data
1491 : * value, while in non-CSV mode, \. cannot be a data value.
1492 : */
1493 24745408 : not_end_of_copy:
1494 24753402 : first_char_in_line = false;
1495 : } /* end of outer loop */
1496 :
1497 : /*
1498 : * Transfer any still-uncopied data to line_buf.
1499 : */
1500 1250240 : REFILL_LINEBUF;
1501 :
1502 1250240 : return result;
1503 : }
1504 :
1505 : /*
1506 : * Return decimal value for a hexadecimal digit
1507 : */
1508 : static int
1509 0 : GetDecimalFromHex(char hex)
1510 : {
1511 0 : if (isdigit((unsigned char) hex))
1512 0 : return hex - '0';
1513 : else
1514 0 : return tolower((unsigned char) hex) - 'a' + 10;
1515 : }
1516 :
1517 : /*
1518 : * Parse the current line into separate attributes (fields),
1519 : * performing de-escaping as needed.
1520 : *
1521 : * The input is in line_buf. We use attribute_buf to hold the result
1522 : * strings. cstate->raw_fields[k] is set to point to the k'th attribute
1523 : * string, or NULL when the input matches the null marker string.
1524 : * This array is expanded as necessary.
1525 : *
1526 : * (Note that the caller cannot check for nulls since the returned
1527 : * string would be the post-de-escaping equivalent, which may look
1528 : * the same as some valid data string.)
1529 : *
1530 : * delim is the column delimiter string (must be just one byte for now).
1531 : * null_print is the null marker string. Note that this is compared to
1532 : * the pre-de-escaped input string.
1533 : *
1534 : * The return value is the number of fields actually read.
1535 : */
1536 : static int
1537 1248398 : CopyReadAttributesText(CopyFromState cstate)
1538 : {
1539 1248398 : char delimc = cstate->opts.delim[0];
1540 : int fieldno;
1541 : char *output_ptr;
1542 : char *cur_ptr;
1543 : char *line_end_ptr;
1544 :
1545 : /*
1546 : * We need a special case for zero-column tables: check that the input
1547 : * line is empty, and return.
1548 : */
1549 1248398 : if (cstate->max_fields <= 0)
1550 : {
1551 8 : if (cstate->line_buf.len != 0)
1552 0 : ereport(ERROR,
1553 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1554 : errmsg("extra data after last expected column")));
1555 8 : return 0;
1556 : }
1557 :
1558 1248390 : resetStringInfo(&cstate->attribute_buf);
1559 :
1560 : /*
1561 : * The de-escaped attributes will certainly not be longer than the input
1562 : * data line, so we can just force attribute_buf to be large enough and
1563 : * then transfer data without any checks for enough space. We need to do
1564 : * it this way because enlarging attribute_buf mid-stream would invalidate
1565 : * pointers already stored into cstate->raw_fields[].
1566 : */
1567 1248390 : if (cstate->attribute_buf.maxlen <= cstate->line_buf.len)
1568 8 : enlargeStringInfo(&cstate->attribute_buf, cstate->line_buf.len);
1569 1248390 : output_ptr = cstate->attribute_buf.data;
1570 :
1571 : /* set pointer variables for loop */
1572 1248390 : cur_ptr = cstate->line_buf.data;
1573 1248390 : line_end_ptr = cstate->line_buf.data + cstate->line_buf.len;
1574 :
1575 : /* Outer loop iterates over fields */
1576 1248390 : fieldno = 0;
1577 : for (;;)
1578 3179490 : {
1579 4427880 : bool found_delim = false;
1580 : char *start_ptr;
1581 : char *end_ptr;
1582 : int input_len;
1583 4427880 : bool saw_non_ascii = false;
1584 :
1585 : /* Make sure there is enough space for the next value */
1586 4427880 : if (fieldno >= cstate->max_fields)
1587 : {
1588 36 : cstate->max_fields *= 2;
1589 36 : cstate->raw_fields =
1590 36 : repalloc(cstate->raw_fields, cstate->max_fields * sizeof(char *));
1591 : }
1592 :
1593 : /* Remember start of field on both input and output sides */
1594 4427880 : start_ptr = cur_ptr;
1595 4427880 : cstate->raw_fields[fieldno] = output_ptr;
1596 :
1597 : /*
1598 : * Scan data for field.
1599 : *
1600 : * Note that in this loop, we are scanning to locate the end of field
1601 : * and also speculatively performing de-escaping. Once we find the
1602 : * end-of-field, we can match the raw field contents against the null
1603 : * marker string. Only after that comparison fails do we know that
1604 : * de-escaping is actually the right thing to do; therefore we *must
1605 : * not* throw any syntax errors before we've done the null-marker
1606 : * check.
1607 : */
1608 : for (;;)
1609 21569108 : {
1610 : char c;
1611 :
1612 25996988 : end_ptr = cur_ptr;
1613 25996988 : if (cur_ptr >= line_end_ptr)
1614 1248384 : break;
1615 24748604 : c = *cur_ptr++;
1616 24748604 : if (c == delimc)
1617 : {
1618 3179496 : found_delim = true;
1619 3179496 : break;
1620 : }
1621 21569108 : if (c == '\\')
1622 : {
1623 7988 : if (cur_ptr >= line_end_ptr)
1624 0 : break;
1625 7988 : c = *cur_ptr++;
1626 7988 : switch (c)
1627 : {
1628 12 : case '0':
1629 : case '1':
1630 : case '2':
1631 : case '3':
1632 : case '4':
1633 : case '5':
1634 : case '6':
1635 : case '7':
1636 : {
1637 : /* handle \013 */
1638 : int val;
1639 :
1640 12 : val = OCTVALUE(c);
1641 12 : if (cur_ptr < line_end_ptr)
1642 : {
1643 6 : c = *cur_ptr;
1644 6 : if (ISOCTAL(c))
1645 : {
1646 0 : cur_ptr++;
1647 0 : val = (val << 3) + OCTVALUE(c);
1648 0 : if (cur_ptr < line_end_ptr)
1649 : {
1650 0 : c = *cur_ptr;
1651 0 : if (ISOCTAL(c))
1652 : {
1653 0 : cur_ptr++;
1654 0 : val = (val << 3) + OCTVALUE(c);
1655 : }
1656 : }
1657 : }
1658 : }
1659 12 : c = val & 0377;
1660 12 : if (c == '\0' || IS_HIGHBIT_SET(c))
1661 12 : saw_non_ascii = true;
1662 : }
1663 12 : break;
1664 12 : case 'x':
1665 : /* Handle \x3F */
1666 12 : if (cur_ptr < line_end_ptr)
1667 : {
1668 6 : char hexchar = *cur_ptr;
1669 :
1670 6 : if (isxdigit((unsigned char) hexchar))
1671 : {
1672 0 : int val = GetDecimalFromHex(hexchar);
1673 :
1674 0 : cur_ptr++;
1675 0 : if (cur_ptr < line_end_ptr)
1676 : {
1677 0 : hexchar = *cur_ptr;
1678 0 : if (isxdigit((unsigned char) hexchar))
1679 : {
1680 0 : cur_ptr++;
1681 0 : val = (val << 4) + GetDecimalFromHex(hexchar);
1682 : }
1683 : }
1684 0 : c = val & 0xff;
1685 0 : if (c == '\0' || IS_HIGHBIT_SET(c))
1686 0 : saw_non_ascii = true;
1687 : }
1688 : }
1689 12 : break;
1690 0 : case 'b':
1691 0 : c = '\b';
1692 0 : break;
1693 0 : case 'f':
1694 0 : c = '\f';
1695 0 : break;
1696 3050 : case 'n':
1697 3050 : c = '\n';
1698 3050 : break;
1699 0 : case 'r':
1700 0 : c = '\r';
1701 0 : break;
1702 0 : case 't':
1703 0 : c = '\t';
1704 0 : break;
1705 0 : case 'v':
1706 0 : c = '\v';
1707 0 : break;
1708 :
1709 : /*
1710 : * in all other cases, take the char after '\'
1711 : * literally
1712 : */
1713 : }
1714 21561120 : }
1715 :
1716 : /* Add c to output string */
1717 21569108 : *output_ptr++ = c;
1718 : }
1719 :
1720 : /* Check whether raw input matched null marker */
1721 4427880 : input_len = end_ptr - start_ptr;
1722 4427880 : if (input_len == cstate->opts.null_print_len &&
1723 241420 : strncmp(start_ptr, cstate->opts.null_print, input_len) == 0)
1724 4802 : cstate->raw_fields[fieldno] = NULL;
1725 : /* Check whether raw input matched default marker */
1726 4423078 : else if (fieldno < list_length(cstate->attnumlist) &&
1727 4423036 : cstate->opts.default_print &&
1728 114 : input_len == cstate->opts.default_print_len &&
1729 30 : strncmp(start_ptr, cstate->opts.default_print, input_len) == 0)
1730 24 : {
1731 : /* fieldno is 0-indexed and attnum is 1-indexed */
1732 30 : int m = list_nth_int(cstate->attnumlist, fieldno) - 1;
1733 :
1734 30 : if (cstate->defexprs[m] != NULL)
1735 : {
1736 : /* defaults contain entries for all physical attributes */
1737 24 : cstate->defaults[m] = true;
1738 : }
1739 : else
1740 : {
1741 6 : TupleDesc tupDesc = RelationGetDescr(cstate->rel);
1742 6 : Form_pg_attribute att = TupleDescAttr(tupDesc, m);
1743 :
1744 6 : ereport(ERROR,
1745 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1746 : errmsg("unexpected default marker in COPY data"),
1747 : errdetail("Column \"%s\" has no default value.",
1748 : NameStr(att->attname))));
1749 : }
1750 : }
1751 : else
1752 : {
1753 : /*
1754 : * At this point we know the field is supposed to contain data.
1755 : *
1756 : * If we de-escaped any non-7-bit-ASCII chars, make sure the
1757 : * resulting string is valid data for the db encoding.
1758 : */
1759 4423048 : if (saw_non_ascii)
1760 : {
1761 0 : char *fld = cstate->raw_fields[fieldno];
1762 :
1763 0 : pg_verifymbstr(fld, output_ptr - fld, false);
1764 : }
1765 : }
1766 :
1767 : /* Terminate attribute value in output area */
1768 4427874 : *output_ptr++ = '\0';
1769 :
1770 4427874 : fieldno++;
1771 : /* Done if we hit EOL instead of a delim */
1772 4427874 : if (!found_delim)
1773 1248384 : break;
1774 : }
1775 :
1776 : /* Clean up state of attribute_buf */
1777 1248384 : output_ptr--;
1778 : Assert(*output_ptr == '\0');
1779 1248384 : cstate->attribute_buf.len = (output_ptr - cstate->attribute_buf.data);
1780 :
1781 1248384 : return fieldno;
1782 : }
1783 :
1784 : /*
1785 : * Parse the current line into separate attributes (fields),
1786 : * performing de-escaping as needed. This has exactly the same API as
1787 : * CopyReadAttributesText, except we parse the fields according to
1788 : * "standard" (i.e. common) CSV usage.
1789 : */
1790 : static int
1791 404 : CopyReadAttributesCSV(CopyFromState cstate)
1792 : {
1793 404 : char delimc = cstate->opts.delim[0];
1794 404 : char quotec = cstate->opts.quote[0];
1795 404 : char escapec = cstate->opts.escape[0];
1796 : int fieldno;
1797 : char *output_ptr;
1798 : char *cur_ptr;
1799 : char *line_end_ptr;
1800 :
1801 : /*
1802 : * We need a special case for zero-column tables: check that the input
1803 : * line is empty, and return.
1804 : */
1805 404 : if (cstate->max_fields <= 0)
1806 : {
1807 0 : if (cstate->line_buf.len != 0)
1808 0 : ereport(ERROR,
1809 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1810 : errmsg("extra data after last expected column")));
1811 0 : return 0;
1812 : }
1813 :
1814 404 : resetStringInfo(&cstate->attribute_buf);
1815 :
1816 : /*
1817 : * The de-escaped attributes will certainly not be longer than the input
1818 : * data line, so we can just force attribute_buf to be large enough and
1819 : * then transfer data without any checks for enough space. We need to do
1820 : * it this way because enlarging attribute_buf mid-stream would invalidate
1821 : * pointers already stored into cstate->raw_fields[].
1822 : */
1823 404 : if (cstate->attribute_buf.maxlen <= cstate->line_buf.len)
1824 0 : enlargeStringInfo(&cstate->attribute_buf, cstate->line_buf.len);
1825 404 : output_ptr = cstate->attribute_buf.data;
1826 :
1827 : /* set pointer variables for loop */
1828 404 : cur_ptr = cstate->line_buf.data;
1829 404 : line_end_ptr = cstate->line_buf.data + cstate->line_buf.len;
1830 :
1831 : /* Outer loop iterates over fields */
1832 404 : fieldno = 0;
1833 : for (;;)
1834 490 : {
1835 894 : bool found_delim = false;
1836 894 : bool saw_quote = false;
1837 : char *start_ptr;
1838 : char *end_ptr;
1839 : int input_len;
1840 :
1841 : /* Make sure there is enough space for the next value */
1842 894 : if (fieldno >= cstate->max_fields)
1843 : {
1844 0 : cstate->max_fields *= 2;
1845 0 : cstate->raw_fields =
1846 0 : repalloc(cstate->raw_fields, cstate->max_fields * sizeof(char *));
1847 : }
1848 :
1849 : /* Remember start of field on both input and output sides */
1850 894 : start_ptr = cur_ptr;
1851 894 : cstate->raw_fields[fieldno] = output_ptr;
1852 :
1853 : /*
1854 : * Scan data for field,
1855 : *
1856 : * The loop starts in "not quote" mode and then toggles between that
1857 : * and "in quote" mode. The loop exits normally if it is in "not
1858 : * quote" mode and a delimiter or line end is seen.
1859 : */
1860 : for (;;)
1861 182 : {
1862 : char c;
1863 :
1864 : /* Not in quote */
1865 : for (;;)
1866 : {
1867 2866 : end_ptr = cur_ptr;
1868 2866 : if (cur_ptr >= line_end_ptr)
1869 398 : goto endfield;
1870 2468 : c = *cur_ptr++;
1871 : /* unquoted field delimiter */
1872 2468 : if (c == delimc)
1873 : {
1874 496 : found_delim = true;
1875 496 : goto endfield;
1876 : }
1877 : /* start of quoted field (or part of field) */
1878 1972 : if (c == quotec)
1879 : {
1880 182 : saw_quote = true;
1881 182 : break;
1882 : }
1883 : /* Add c to output string */
1884 1790 : *output_ptr++ = c;
1885 : }
1886 :
1887 : /* In quote */
1888 : for (;;)
1889 : {
1890 1170 : end_ptr = cur_ptr;
1891 1170 : if (cur_ptr >= line_end_ptr)
1892 0 : ereport(ERROR,
1893 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1894 : errmsg("unterminated CSV quoted field")));
1895 :
1896 1170 : c = *cur_ptr++;
1897 :
1898 : /* escape within a quoted field */
1899 1170 : if (c == escapec)
1900 : {
1901 : /*
1902 : * peek at the next char if available, and escape it if it
1903 : * is an escape char or a quote char
1904 : */
1905 118 : if (cur_ptr < line_end_ptr)
1906 : {
1907 72 : char nextc = *cur_ptr;
1908 :
1909 72 : if (nextc == escapec || nextc == quotec)
1910 : {
1911 24 : *output_ptr++ = nextc;
1912 24 : cur_ptr++;
1913 24 : continue;
1914 : }
1915 : }
1916 : }
1917 :
1918 : /*
1919 : * end of quoted field. Must do this test after testing for
1920 : * escape in case quote char and escape char are the same
1921 : * (which is the common case).
1922 : */
1923 1146 : if (c == quotec)
1924 182 : break;
1925 :
1926 : /* Add c to output string */
1927 964 : *output_ptr++ = c;
1928 : }
1929 : }
1930 894 : endfield:
1931 :
1932 : /* Terminate attribute value in output area */
1933 894 : *output_ptr++ = '\0';
1934 :
1935 : /* Check whether raw input matched null marker */
1936 894 : input_len = end_ptr - start_ptr;
1937 894 : if (!saw_quote && input_len == cstate->opts.null_print_len &&
1938 44 : strncmp(start_ptr, cstate->opts.null_print, input_len) == 0)
1939 44 : cstate->raw_fields[fieldno] = NULL;
1940 : /* Check whether raw input matched default marker */
1941 850 : else if (fieldno < list_length(cstate->attnumlist) &&
1942 850 : cstate->opts.default_print &&
1943 150 : input_len == cstate->opts.default_print_len &&
1944 42 : strncmp(start_ptr, cstate->opts.default_print, input_len) == 0)
1945 : {
1946 : /* fieldno is 0-index and attnum is 1-index */
1947 42 : int m = list_nth_int(cstate->attnumlist, fieldno) - 1;
1948 :
1949 42 : if (cstate->defexprs[m] != NULL)
1950 : {
1951 : /* defaults contain entries for all physical attributes */
1952 36 : cstate->defaults[m] = true;
1953 : }
1954 : else
1955 : {
1956 6 : TupleDesc tupDesc = RelationGetDescr(cstate->rel);
1957 6 : Form_pg_attribute att = TupleDescAttr(tupDesc, m);
1958 :
1959 6 : ereport(ERROR,
1960 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1961 : errmsg("unexpected default marker in COPY data"),
1962 : errdetail("Column \"%s\" has no default value.",
1963 : NameStr(att->attname))));
1964 : }
1965 : }
1966 :
1967 888 : fieldno++;
1968 : /* Done if we hit EOL instead of a delim */
1969 888 : if (!found_delim)
1970 398 : break;
1971 : }
1972 :
1973 : /* Clean up state of attribute_buf */
1974 398 : output_ptr--;
1975 : Assert(*output_ptr == '\0');
1976 398 : cstate->attribute_buf.len = (output_ptr - cstate->attribute_buf.data);
1977 :
1978 398 : return fieldno;
1979 : }
1980 :
1981 :
1982 : /*
1983 : * Read a binary attribute
1984 : */
1985 : static Datum
1986 158 : CopyReadBinaryAttribute(CopyFromState cstate, FmgrInfo *flinfo,
1987 : Oid typioparam, int32 typmod,
1988 : bool *isnull)
1989 : {
1990 : int32 fld_size;
1991 : Datum result;
1992 :
1993 158 : if (!CopyGetInt32(cstate, &fld_size))
1994 0 : ereport(ERROR,
1995 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
1996 : errmsg("unexpected EOF in COPY data")));
1997 158 : if (fld_size == -1)
1998 : {
1999 30 : *isnull = true;
2000 30 : return ReceiveFunctionCall(flinfo, NULL, typioparam, typmod);
2001 : }
2002 128 : if (fld_size < 0)
2003 0 : ereport(ERROR,
2004 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
2005 : errmsg("invalid field size")));
2006 :
2007 : /* reset attribute_buf to empty, and load raw data in it */
2008 128 : resetStringInfo(&cstate->attribute_buf);
2009 :
2010 128 : enlargeStringInfo(&cstate->attribute_buf, fld_size);
2011 128 : if (CopyReadBinaryData(cstate, cstate->attribute_buf.data,
2012 128 : fld_size) != fld_size)
2013 0 : ereport(ERROR,
2014 : (errcode(ERRCODE_BAD_COPY_FILE_FORMAT),
2015 : errmsg("unexpected EOF in COPY data")));
2016 :
2017 128 : cstate->attribute_buf.len = fld_size;
2018 128 : cstate->attribute_buf.data[fld_size] = '\0';
2019 :
2020 : /* Call the column type's binary input converter */
2021 128 : result = ReceiveFunctionCall(flinfo, &cstate->attribute_buf,
2022 : typioparam, typmod);
2023 :
2024 : /* Trouble if it didn't eat the whole buffer */
2025 128 : if (cstate->attribute_buf.cursor != cstate->attribute_buf.len)
2026 2 : ereport(ERROR,
2027 : (errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
2028 : errmsg("incorrect binary data format")));
2029 :
2030 126 : *isnull = false;
2031 126 : return result;
2032 : }
|
