1 /*-------------------------------------------------------------------------
5 * Implements the custom output format.
7 * The comments with the routined in this code are a good place to
8 * understand how to write a new format.
10 * See the headers to pg_restore for more details.
12 * Copyright (c) 2000, Philip Warner
13 * Rights are granted to use this software in any way so long
14 * as this notice is not removed.
16 * The author is not responsible for loss or damages that may
17 * and any liability will be limited to the time taken to fix any
22 * $Header: /cvsroot/pgsql/src/bin/pg_dump/pg_backup_custom.c,v 1.21 2002/09/04 20:31:34 momjian Exp $
24 *-------------------------------------------------------------------------
27 #include "pg_backup.h"
28 #include "pg_backup_archiver.h"
33 * Routines in the format interface
37 static void _ArchiveEntry(ArchiveHandle *AH, TocEntry *te);
38 static void _StartData(ArchiveHandle *AH, TocEntry *te);
39 static size_t _WriteData(ArchiveHandle *AH, const void *data, size_t dLen);
40 static void _EndData(ArchiveHandle *AH, TocEntry *te);
41 static int _WriteByte(ArchiveHandle *AH, const int i);
42 static int _ReadByte(ArchiveHandle *);
43 static size_t _WriteBuf(ArchiveHandle *AH, const void *buf, size_t len);
44 static size_t _ReadBuf(ArchiveHandle *AH, void *buf, size_t len);
45 static void _CloseArchive(ArchiveHandle *AH);
46 static void _PrintTocData(ArchiveHandle *AH, TocEntry *te, RestoreOptions *ropt);
47 static void _WriteExtraToc(ArchiveHandle *AH, TocEntry *te);
48 static void _ReadExtraToc(ArchiveHandle *AH, TocEntry *te);
49 static void _PrintExtraToc(ArchiveHandle *AH, TocEntry *te);
51 static void _PrintData(ArchiveHandle *AH);
52 static void _skipData(ArchiveHandle *AH);
53 static void _skipBlobs(ArchiveHandle *AH);
55 static void _StartBlobs(ArchiveHandle *AH, TocEntry *te);
56 static void _StartBlob(ArchiveHandle *AH, TocEntry *te, Oid oid);
57 static void _EndBlob(ArchiveHandle *AH, TocEntry *te, Oid oid);
58 static void _EndBlobs(ArchiveHandle *AH, TocEntry *te);
59 static void _LoadBlobs(ArchiveHandle *AH);
62 * Buffers used in zlib compression and extra data stored in archive and
66 #define zlibOutSize 4096
67 #define zlibInSize 4096
91 static void _readBlockHeader(ArchiveHandle *AH, int *type, int *id);
92 static void _StartDataCompressor(ArchiveHandle *AH, TocEntry *te);
93 static void _EndDataCompressor(ArchiveHandle *AH, TocEntry *te);
94 static off_t _getFilePos(ArchiveHandle *AH, lclContext *ctx);
95 static int _DoDeflate(ArchiveHandle *AH, lclContext *ctx, int flush);
97 static char *modulename = gettext_noop("custom archiver");
102 * Init routine required by ALL formats. This is a global routine
103 * and should be declared in pg_backup_archiver.h
105 * It's task is to create any extra archive context (using AH->formatData),
106 * and to initialize the supported function pointers.
108 * It should also prepare whatever it's input source is for reading/writing,
109 * and in the case of a read mode connection, it should load the Header & TOC.
112 InitArchiveFmt_Custom(ArchiveHandle *AH)
116 /* Assuming static functions, this can be copied for each format. */
117 AH->ArchiveEntryPtr = _ArchiveEntry;
118 AH->StartDataPtr = _StartData;
119 AH->WriteDataPtr = _WriteData;
120 AH->EndDataPtr = _EndData;
121 AH->WriteBytePtr = _WriteByte;
122 AH->ReadBytePtr = _ReadByte;
123 AH->WriteBufPtr = _WriteBuf;
124 AH->ReadBufPtr = _ReadBuf;
125 AH->ClosePtr = _CloseArchive;
126 AH->PrintTocDataPtr = _PrintTocData;
127 AH->ReadExtraTocPtr = _ReadExtraToc;
128 AH->WriteExtraTocPtr = _WriteExtraToc;
129 AH->PrintExtraTocPtr = _PrintExtraToc;
131 AH->StartBlobsPtr = _StartBlobs;
132 AH->StartBlobPtr = _StartBlob;
133 AH->EndBlobPtr = _EndBlob;
134 AH->EndBlobsPtr = _EndBlobs;
137 * Set up some special context used in compressing data.
139 ctx = (lclContext *) malloc(sizeof(lclContext));
141 die_horribly(AH, modulename, "out of memory\n");
142 AH->formatData = (void *) ctx;
144 ctx->zp = (z_streamp) malloc(sizeof(z_stream));
146 die_horribly(AH, modulename, "out of memory\n");
148 /* Initialize LO buffering */
149 AH->lo_buf_size = LOBBUFSIZE;
150 AH->lo_buf = (void *) malloc(LOBBUFSIZE);
151 if (AH->lo_buf == NULL)
152 die_horribly(AH, modulename, "out of memory\n");
155 * zlibOutSize is the buffer size we tell zlib it can output to. We
156 * actually allocate one extra byte because some routines want to
157 * append a trailing zero byte to the zlib output. The input buffer
158 * is expansible and is always of size ctx->inSize; zlibInSize is just
159 * the initial default size for it.
161 ctx->zlibOut = (char *) malloc(zlibOutSize + 1);
162 ctx->zlibIn = (char *) malloc(zlibInSize);
163 ctx->inSize = zlibInSize;
166 if (ctx->zlibOut == NULL || ctx->zlibIn == NULL)
167 die_horribly(AH, modulename, "out of memory\n");
172 if (AH->mode == archModeWrite)
175 if (AH->fSpec && strcmp(AH->fSpec, "") != 0)
176 AH->FH = fopen(AH->fSpec, PG_BINARY_W);
181 die_horribly(AH, modulename, "could not open archive file %s: %s\n", AH->fSpec, strerror(errno));
183 ctx->hasSeek = (fseeko(AH->FH, 0, SEEK_CUR) == 0);
189 if (AH->fSpec && strcmp(AH->fSpec, "") != 0)
190 AH->FH = fopen(AH->fSpec, PG_BINARY_R);
194 die_horribly(AH, modulename, "could not open archive file %s: %s\n", AH->fSpec, strerror(errno));
196 ctx->hasSeek = (fseeko(AH->FH, 0, SEEK_CUR) == 0);
200 ctx->dataStart = _getFilePos(AH, ctx);
206 * Called by the Archiver when the dumper creates a new TOC entry.
210 * Set up extrac format-related TOC data.
213 _ArchiveEntry(ArchiveHandle *AH, TocEntry *te)
217 ctx = (lclTocEntry *) calloc(1, sizeof(lclTocEntry));
223 te->formatData = (void *) ctx;
228 * Called by the Archiver to save any extra format-related TOC entry
233 * Use the Archiver routines to write data - they are non-endian, and
234 * maintain other important file information.
237 _WriteExtraToc(ArchiveHandle *AH, TocEntry *te)
239 lclTocEntry *ctx = (lclTocEntry *) te->formatData;
241 WriteInt(AH, ctx->dataPos);
242 WriteInt(AH, ctx->dataLen);
246 * Called by the Archiver to read any extra format-related TOC data.
250 * Needs to match the order defined in _WriteExtraToc, and sould also
251 * use the Archiver input routines.
254 _ReadExtraToc(ArchiveHandle *AH, TocEntry *te)
256 lclTocEntry *ctx = (lclTocEntry *) te->formatData;
260 ctx = (lclTocEntry *) malloc(sizeof(lclTocEntry));
261 te->formatData = (void *) ctx;
264 ctx->dataPos = ReadInt(AH);
265 ctx->dataLen = ReadInt(AH);
269 * Called by the Archiver when restoring an archive to output a comment
270 * that includes useful information about the TOC entry.
276 _PrintExtraToc(ArchiveHandle *AH, TocEntry *te)
278 lclTocEntry *ctx = (lclTocEntry *) te->formatData;
280 ahprintf(AH, "-- Data Pos: " INT64_FORMAT " (Length %lu)\n",
281 (int64) ctx->dataPos, (unsigned long) ctx->dataLen);
285 * Called by the archiver when saving TABLE DATA (not schema). This routine
286 * should save whatever format-specific information is needed to read
289 * It is called just prior to the dumper's 'DataDumper' routine being called.
291 * Optional, but strongly recommended.
295 _StartData(ArchiveHandle *AH, TocEntry *te)
297 lclContext *ctx = (lclContext *) AH->formatData;
298 lclTocEntry *tctx = (lclTocEntry *) te->formatData;
300 tctx->dataPos = _getFilePos(AH, ctx);
302 _WriteByte(AH, BLK_DATA); /* Block type */
303 WriteInt(AH, te->id); /* For sanity check */
305 _StartDataCompressor(AH, te);
310 * Called by archiver when dumper calls WriteData. This routine is
311 * called for both BLOB and TABLE data; it is the responsibility of
312 * the format to manage each kind of data using StartBlob/StartData.
314 * It should only be called from withing a DataDumper routine.
320 _WriteData(ArchiveHandle *AH, const void *data, size_t dLen)
322 lclContext *ctx = (lclContext *) AH->formatData;
323 z_streamp zp = ctx->zp;
325 zp->next_in = (void *) data;
328 while (zp->avail_in != 0)
330 /* printf("Deflating %lu bytes\n", (unsigned long) dLen); */
331 _DoDeflate(AH, ctx, 0);
337 * Called by the archiver when a dumper's 'DataDumper' routine has
344 _EndData(ArchiveHandle *AH, TocEntry *te)
346 lclContext *ctx = (lclContext *) AH->formatData;
347 lclTocEntry *tctx = (lclTocEntry *) te->formatData;
349 _EndDataCompressor(AH, te);
351 tctx->dataLen = _getFilePos(AH, ctx) - tctx->dataPos;
355 * Called by the archiver when starting to save all BLOB DATA (not schema).
356 * This routine should save whatever format-specific information is needed
357 * to read the BLOBs back into memory.
359 * It is called just prior to the dumper's DataDumper routine.
361 * Optional, but strongly recommended.
365 _StartBlobs(ArchiveHandle *AH, TocEntry *te)
367 lclContext *ctx = (lclContext *) AH->formatData;
368 lclTocEntry *tctx = (lclTocEntry *) te->formatData;
370 tctx->dataPos = _getFilePos(AH, ctx);
372 _WriteByte(AH, BLK_BLOBS); /* Block type */
373 WriteInt(AH, te->id); /* For sanity check */
378 * Called by the archiver when the dumper calls StartBlob.
382 * Must save the passed OID for retrieval at restore-time.
385 _StartBlob(ArchiveHandle *AH, TocEntry *te, Oid oid)
388 die_horribly(AH, modulename, "invalid OID for large object\n");
391 _StartDataCompressor(AH, te);
395 * Called by the archiver when the dumper calls EndBlob.
401 _EndBlob(ArchiveHandle *AH, TocEntry *te, Oid oid)
403 _EndDataCompressor(AH, te);
407 * Called by the archiver when finishing saving all BLOB DATA.
413 _EndBlobs(ArchiveHandle *AH, TocEntry *te)
415 /* Write out a fake zero OID to mark end-of-blobs. */
420 * Print data for a gievn TOC entry
423 _PrintTocData(ArchiveHandle *AH, TocEntry *te, RestoreOptions *ropt)
425 lclContext *ctx = (lclContext *) AH->formatData;
427 lclTocEntry *tctx = (lclTocEntry *) te->formatData;
431 if (tctx->dataPos == 0)
434 if (!ctx->hasSeek || tctx->dataPos < 0)
437 /* Skip over unnecessary blocks until we get the one we want. */
441 _readBlockHeader(AH, &blkType, &id);
446 if ((TocIDRequired(AH, id, ropt) & 2) != 0)
447 die_horribly(AH, modulename,
448 "Dumping a specific TOC data block out of order is not supported"
449 " without id on this input stream (fseek required)\n");
464 default: /* Always have a default */
466 die_horribly(AH, modulename,
467 "unrecognized data block type (%d) while searching archive\n",
472 _readBlockHeader(AH, &blkType, &id);
482 if (fseeko(AH->FH, tctx->dataPos, SEEK_SET) != 0)
483 die_horribly(AH, modulename, "error during file seek: %s\n", strerror(errno));
485 _readBlockHeader(AH, &blkType, &id);
491 die_horribly(AH, modulename, "found unexpected block ID (%d) when reading data - expected %d\n",
505 die_horribly(AH, modulename, "large objects cannot be loaded without a database connection\n");
510 default: /* Always have a default */
512 die_horribly(AH, modulename, "unrecognized data block type %d while restoring archive\n",
517 ahprintf(AH, "\n\n");
521 * Print data from current file position.
524 _PrintData(ArchiveHandle *AH)
526 lclContext *ctx = (lclContext *) AH->formatData;
527 z_streamp zp = ctx->zp;
529 char *in = ctx->zlibIn;
534 char *out = ctx->zlibOut;
541 if (AH->compression != 0)
547 if (inflateInit(zp) != Z_OK)
548 die_horribly(AH, modulename, "could not initialize compression library: %s\n", zp->msg);
552 blkLen = ReadInt(AH);
555 if (blkLen + 1 > ctx->inSize)
559 ctx->zlibIn = (char *) malloc(blkLen + 1);
561 die_horribly(AH, modulename, "out of memory\n");
563 ctx->inSize = blkLen + 1;
567 cnt = fread(in, 1, blkLen, AH->FH);
569 die_horribly(AH, modulename,
570 "could not read data block - expected %lu, got %lu\n",
571 (unsigned long) blkLen, (unsigned long) cnt);
573 ctx->filePos += blkLen;
576 zp->avail_in = blkLen;
580 if (AH->compression != 0)
583 while (zp->avail_in != 0)
586 zp->avail_out = zlibOutSize;
587 res = inflate(zp, 0);
588 if (res != Z_OK && res != Z_STREAM_END)
589 die_horribly(AH, modulename, "unable to uncompress data: %s\n", zp->msg);
591 out[zlibOutSize - zp->avail_out] = '\0';
592 ahwrite(out, 1, zlibOutSize - zp->avail_out, AH);
598 in[zp->avail_in] = '\0';
599 ahwrite(in, 1, zp->avail_in, AH);
606 blkLen = ReadInt(AH);
611 if (AH->compression != 0)
615 while (res != Z_STREAM_END)
618 zp->avail_out = zlibOutSize;
619 res = inflate(zp, 0);
620 if (res != Z_OK && res != Z_STREAM_END)
621 die_horribly(AH, modulename, "unable to uncompress data: %s\n", zp->msg);
623 out[zlibOutSize - zp->avail_out] = '\0';
624 ahwrite(out, 1, zlibOutSize - zp->avail_out, AH);
626 if (inflateEnd(zp) != Z_OK)
627 die_horribly(AH, modulename, "could not close compression library: %s\n", zp->msg);
634 _LoadBlobs(ArchiveHandle *AH)
638 StartRestoreBlobs(AH);
643 StartRestoreBlob(AH, oid);
645 EndRestoreBlob(AH, oid);
654 * Skip the BLOBs from the current file position.
655 * BLOBS are written sequentially as data blocks (see below).
656 * Each BLOB is preceded by it's original OID.
657 * A zero OID indicated the end of the BLOBS
660 _skipBlobs(ArchiveHandle *AH)
673 * Skip data from current file position.
674 * Data blocks are formatted as an integer length, followed by data.
675 * A zero length denoted the end of the block.
678 _skipData(ArchiveHandle *AH)
680 lclContext *ctx = (lclContext *) AH->formatData;
682 char *in = ctx->zlibIn;
685 blkLen = ReadInt(AH);
688 if (blkLen > ctx->inSize)
691 ctx->zlibIn = (char *) malloc(blkLen);
692 ctx->inSize = blkLen;
695 cnt = fread(in, 1, blkLen, AH->FH);
697 die_horribly(AH, modulename,
698 "could not read data block - expected %lu, got %lu\n",
699 (unsigned long) blkLen, (unsigned long) cnt);
701 ctx->filePos += blkLen;
703 blkLen = ReadInt(AH);
709 * Write a byte of data to the archive.
713 * Called by the archiver to do integer & byte output to the archive.
714 * These routines are only used to read & write headers & TOC.
718 _WriteByte(ArchiveHandle *AH, const int i)
720 lclContext *ctx = (lclContext *) AH->formatData;
723 res = fputc(i, AH->FH);
727 die_horribly(AH, modulename, "could not write byte: %s\n", strerror(errno));
732 * Read a byte of data from the archive.
736 * Called by the archiver to read bytes & integers from the archive.
737 * These routines are only used to read & write headers & TOC.
741 _ReadByte(ArchiveHandle *AH)
743 lclContext *ctx = (lclContext *) AH->formatData;
753 * Write a buffer of data to the archive.
757 * Called by the archiver to write a block of bytes to the archive.
758 * These routines are only used to read & write headers & TOC.
762 _WriteBuf(ArchiveHandle *AH, const void *buf, size_t len)
764 lclContext *ctx = (lclContext *) AH->formatData;
767 res = fwrite(buf, 1, len, AH->FH);
770 die_horribly(AH, modulename,
771 "write error in _WriteBuf (%lu != %lu)\n",
772 (unsigned long) res, (unsigned long) len);
779 * Read a block of bytes from the archive.
783 * Called by the archiver to read a block of bytes from the archive
784 * These routines are only used to read & write headers & TOC.
788 _ReadBuf(ArchiveHandle *AH, void *buf, size_t len)
790 lclContext *ctx = (lclContext *) AH->formatData;
793 res = fread(buf, 1, len, AH->FH);
804 * When writing the archive, this is the routine that actually starts
805 * the process of saving it to files. No data should be written prior
806 * to this point, since the user could sort the TOC after creating it.
808 * If an archive is to be written, this toutine must call:
809 * WriteHead to save the archive header
810 * WriteToc to save the TOC entries
811 * WriteDataChunks to save all DATA & BLOBs.
815 _CloseArchive(ArchiveHandle *AH)
817 lclContext *ctx = (lclContext *) AH->formatData;
820 if (AH->mode == archModeWrite)
823 tpos = ftello(AH->FH);
825 ctx->dataStart = _getFilePos(AH, ctx);
829 * This is not an essential operation - it is really only needed
830 * if we expect to be doing seeks to read the data back - it may
831 * be ok to just use the existing self-consistent block
836 fseeko(AH->FH, tpos, SEEK_SET);
841 if (fclose(AH->FH) != 0)
842 die_horribly(AH, modulename, "could not close archive file: %s\n", strerror(errno));
847 /*--------------------------------------------------
848 * END OF FORMAT CALLBACKS
849 *--------------------------------------------------
853 * Get the current position in the archive file.
856 _getFilePos(ArchiveHandle *AH, lclContext *ctx)
862 pos = ftello(AH->FH);
863 if (pos != ctx->filePos)
865 write_msg(modulename, "WARNING: ftell mismatch with expected position -- ftell ignored\n");
875 * Read a data block header. The format changed in V1.3, so we
876 * put the code here for simplicity.
879 _readBlockHeader(ArchiveHandle *AH, int *type, int *id)
881 if (AH->version < K_VERS_1_3)
884 *type = _ReadByte(AH);;
890 * If zlib is available, then startit up. This is called from
891 * StartData & StartBlob. The buffers are setup in the Init routine.
895 _StartDataCompressor(ArchiveHandle *AH, TocEntry *te)
897 lclContext *ctx = (lclContext *) AH->formatData;
898 z_streamp zp = ctx->zp;
902 if (AH->compression < 0 || AH->compression > 9)
903 AH->compression = Z_DEFAULT_COMPRESSION;
905 if (AH->compression != 0)
911 if (deflateInit(zp, AH->compression) != Z_OK)
912 die_horribly(AH, modulename, "could not initialize compression library: %s\n", zp->msg);
920 /* Just be paranoid - maybe End is called after Start, with no Write */
921 zp->next_out = ctx->zlibOut;
922 zp->avail_out = zlibOutSize;
926 * Send compressed data to the output stream (via ahwrite).
927 * Each data chunk is preceded by it's length.
928 * In the case of Z0, or no zlib, just write the raw data.
932 _DoDeflate(ArchiveHandle *AH, lclContext *ctx, int flush)
934 z_streamp zp = ctx->zp;
937 char *out = ctx->zlibOut;
940 if (AH->compression != 0)
942 res = deflate(zp, flush);
943 if (res == Z_STREAM_ERROR)
944 die_horribly(AH, modulename, "could not compress data: %s\n", zp->msg);
946 if (((flush == Z_FINISH) && (zp->avail_out < zlibOutSize))
947 || (zp->avail_out == 0)
948 || (zp->avail_in != 0)
952 * Extra paranoia: avoid zero-length chunks since a zero
953 * length chunk is the EOF marker. This should never happen
956 if (zp->avail_out < zlibOutSize)
959 * printf("Wrote %lu byte deflated chunk\n", (unsigned
960 * long) (zlibOutSize - zp->avail_out));
962 WriteInt(AH, zlibOutSize - zp->avail_out);
963 if (fwrite(out, 1, zlibOutSize - zp->avail_out, AH->FH) != (zlibOutSize - zp->avail_out))
964 die_horribly(AH, modulename, "could not write compressed chunk\n");
965 ctx->filePos += zlibOutSize - zp->avail_out;
968 zp->avail_out = zlibOutSize;
974 if (zp->avail_in > 0)
976 WriteInt(AH, zp->avail_in);
977 if (fwrite(zp->next_in, 1, zp->avail_in, AH->FH) != zp->avail_in)
978 die_horribly(AH, modulename, "could not write uncompressed chunk\n");
979 ctx->filePos += zp->avail_in;
985 if (flush == Z_FINISH)
1002 * Terminate zlib context and flush it's buffers. If no zlib
1007 _EndDataCompressor(ArchiveHandle *AH, TocEntry *te)
1011 lclContext *ctx = (lclContext *) AH->formatData;
1012 z_streamp zp = ctx->zp;
1015 if (AH->compression != 0)
1022 /* printf("Ending data output\n"); */
1023 res = _DoDeflate(AH, ctx, Z_FINISH);
1024 } while (res != Z_STREAM_END);
1026 if (deflateEnd(zp) != Z_OK)
1027 die_horribly(AH, modulename, "could not close compression stream: %s\n", zp->msg);
1031 /* Send the end marker */