From 6aacd903ffe90607bef360b176caa93aa6457f7c Mon Sep 17 00:00:00 2001
From: Moinak Ghosh <moinakg@gmail.com>
Date: Sat, 9 Nov 2013 16:46:19 +0530
Subject: [PATCH] Structured handling of file types. Handling of already
 compressed data based on compression algorithm. Add a few more extension
 types.

---
 adaptive_compress.c        |  6 ++--
 archive/pc_archive.c       |  5 ++--
 bzip2_compress.c           | 16 +++++++----
 libbsc_compress.c          |  8 ++++--
 lz4_compress.c             |  6 ++++
 lzfx_compress.c            |  6 ++++
 lzma_compress.c            |  2 +-
 pcompress.c                |  5 ++--
 ppmd_compress.c            |  3 +-
 utils/phash/extensions.h   |  7 +++--
 utils/phash/extensions.txt |  7 +++--
 utils/phash/phash.c        | 12 ++++----
 utils/phash/phash.h        |  2 +-
 utils/utils.h              | 58 ++++++++++++++++++++++++++------------
 zlib_compress.c            | 13 +++++++++
 15 files changed, 107 insertions(+), 49 deletions(-)

diff --git a/adaptive_compress.c b/adaptive_compress.c
index 60d2463..abf7828 100644
--- a/adaptive_compress.c
+++ b/adaptive_compress.c
@@ -229,14 +229,14 @@ adapt_compress(void *src, uint64_t srclen, void *dst,
 	 * Use PPMd if some percentage of source is 7-bit textual bytes, otherwise
 	 * use Bzip2 or LZMA.
 	 */
-	if (adat->adapt_mode == 2 && (btype & TYPE_BINARY)) {
+	if (adat->adapt_mode == 2 && (PC_TYPE(btype) == TYPE_BINARY)) {
 		rv = lzma_compress(src, srclen, dst, dstlen, level, chdr, btype, adat->lzma_data);
 		if (rv < 0)
 			return (rv);
 		rv = ADAPT_COMPRESS_LZMA;
 		lzma_count++;
 
-	} else if (adat->adapt_mode == 1 && (btype & TYPE_BINARY)) {
+	} else if (adat->adapt_mode == 1 && (PC_TYPE(btype) == TYPE_BINARY)) {
 		rv = bzip2_compress(src, srclen, dst, dstlen, level, chdr, btype, NULL);
 		if (rv < 0)
 			return (rv);
@@ -245,7 +245,7 @@ adapt_compress(void *src, uint64_t srclen, void *dst,
 
 	} else {
 #ifdef ENABLE_PC_LIBBSC
-		if (adat->bsc_data && (btype & TYPE_MARKUP)) {
+		if (adat->bsc_data && PC_SUBTYPE(btype) == TYPE_MARKUP) {
 			rv = libbsc_compress(src, srclen, dst, dstlen, level, chdr, btype, adat->bsc_data);
 			if (rv < 0)
 				return (rv);
diff --git a/archive/pc_archive.c b/archive/pc_archive.c
index 15c7d1c..a8f9f1d 100644
--- a/archive/pc_archive.c
+++ b/archive/pc_archive.c
@@ -942,6 +942,7 @@ archiver_thread_func(void *dat) {
 			continue;
 		}
 
+		typ = TYPE_UNKNOWN;
 		if (archive_entry_filetype(entry) == AE_IFREG) {
 			if ((typ = detect_type_by_ext(fpath, fpathlen)) != TYPE_UNKNOWN)
 				pctx->ctype = typ;
@@ -1248,9 +1249,9 @@ detect_type_by_data(uchar_t *buf, size_t len)
 	if (buf[0] == 0xe9)
 		return (TYPE_BINARY|TYPE_EXE); // MSDOS COM
 	if (U32_P(buf) == TZSHORT)
-		return (TYPE_BINARY|TYPE_BINARY); // Timezone data
+		return (TYPE_BINARY); // Timezone data
 	if (U32_P(buf) == PPMSHORT)
-		return (TYPE_BINARY|TYPE_COMPRESSED); // PPM Compressed archive
+		return (TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_PPMD); // PPM Compressed archive
 
 	return (TYPE_UNKNOWN);
 }
diff --git a/bzip2_compress.c b/bzip2_compress.c
index 4306650..86b4537 100644
--- a/bzip2_compress.c
+++ b/bzip2_compress.c
@@ -174,12 +174,16 @@ bzip2_decompress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
 	char *dst1 = (char *)dst;
 	char *src1 = (char *)src;
 
-	if (btype & TYPE_COMPRESSED) {
-		if ((btype & TYPE_COMPRESSED_LZW) != TYPE_COMPRESSED_LZW &&
-		    (btype & TYPE_COMPRESSED_GZ) != TYPE_COMPRESSED_GZ &&
-		    (btype & TYPE_COMPRESSED_LZ) != TYPE_COMPRESSED_LZ &&
-		    (btype & TYPE_COMPRESSED_LZO) != TYPE_COMPRESSED_LZO)
-		{
+	/*
+	 * If the data is known to be compressed then certain types less compressed data
+	 * can be attempted to be compressed again for a possible gain. For others it is
+	 * a waste of time.
+	 */
+	if (PC_TYPE(btype) == TYPE_COMPRESSED && level < 7) {
+		int subtype = PC_SUBTYPE(btype);
+
+		if (subtype != TYPE_COMPRESSED_LZW && subtype != TYPE_COMPRESSED_GZ &&
+		    subtype != TYPE_COMPRESSED_LZ && subtype != TYPE_COMPRESSED_LZO) {
 			return (-1);
 		}
 	}
diff --git a/libbsc_compress.c b/libbsc_compress.c
index f259096..0eeb104 100644
--- a/libbsc_compress.c
+++ b/libbsc_compress.c
@@ -153,9 +153,11 @@ libbsc_compress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
 	int rv;
 	struct libbsc_params *bscdat = (struct libbsc_params *)data;
 
-	if ((btype & TYPE_COMPRESSED_BZ2) == TYPE_COMPRESSED_BZ2 ||
-	    (btype & TYPE_COMPRESSED_LZMA) == TYPE_COMPRESSED_LZMA)
-		return (-1);
+	if (PC_TYPE(btype) == TYPE_COMPRESSED) {
+		int subtype = PC_SUBTYPE(btype);
+		if (subtype == TYPE_COMPRESSED_BZ2 || subtype == TYPE_COMPRESSED_LZMA)
+			return (-1);
+	}
 
 	rv = bsc_compress(src, dst, srclen, bscdat->lzpHashSize, bscdat->lzpMinLen,
 	    LIBBSC_BLOCKSORTER_BWT, bscdat->bscCoder, bscdat->features);
diff --git a/lz4_compress.c b/lz4_compress.c
index b0e5a36..5ac8574 100644
--- a/lz4_compress.c
+++ b/lz4_compress.c
@@ -106,6 +106,12 @@ lz4_compress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
 	int _srclen = srclen;
 	uchar_t *dst2;
 
+	/*
+	 * Ignore compressed data in fast modes.
+	 */
+	if (lzdat->level < 3 && PC_TYPE(btype) == TYPE_COMPRESSED)
+		return (-1);
+
 	if (lzdat->level == 1) {
 		rv = LZ4_compress((const char *)src, (char *)dst, _srclen);
 
diff --git a/lzfx_compress.c b/lzfx_compress.c
index f687a74..fd6f61e 100644
--- a/lzfx_compress.c
+++ b/lzfx_compress.c
@@ -111,6 +111,12 @@ lz_fx_compress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
 	unsigned int _srclen = srclen;
 	unsigned int _dstlen = *dstlen;
 
+	/*
+	 * Ignore compressed data in fast modes.
+	 */
+	if (level < 7 && PC_TYPE(btype) == TYPE_COMPRESSED)
+		return (-1);
+
 	rv = lzfx_compress(src, _srclen, dst, &_dstlen, lzdat->htab_bits);
 	if (rv != 0) {
 		if (rv != LZFX_ESIZE)
diff --git a/lzma_compress.c b/lzma_compress.c
index 8ddf643..c016820 100644
--- a/lzma_compress.c
+++ b/lzma_compress.c
@@ -211,7 +211,7 @@ lzma_compress(void *src, uint64_t srclen, void *dst,
 		return (-1);
 	}
 
-	if ((btype & TYPE_COMPRESSED_LZMA) == TYPE_COMPRESSED_LZMA)
+	if (PC_SUBTYPE(btype) == TYPE_COMPRESSED_LZMA)
 		return (-1);
 	props->level = level;
 
diff --git a/pcompress.c b/pcompress.c
index 29385b2..242b5cd 100644
--- a/pcompress.c
+++ b/pcompress.c
@@ -1544,12 +1544,13 @@ plain_index:
 		o_chunksize = _chunksize;
 
 		/* Compress data chunk. */
-		if ((pctx->lzp_preprocess || pctx->enable_delta2_encode) && _chunksize > 0) {
+		if ((pctx->lzp_preprocess || pctx->enable_delta2_encode) && _chunksize > 0 &&
+		    PC_SUBTYPE(pctx->btype) == TYPE_CMP_MAX) {
 			rv = preproc_compress(pctx, tdat->compress, tdat->uncompressed_chunk + dedupe_index_sz,
 			    _chunksize, compressed_chunk + index_size_cmp, &_chunksize,
 			    tdat->level, 0, pctx->btype, tdat->data, tdat->props);
 
-		} else if (_chunksize > 0) {
+		} else if (_chunksize > 0 && PC_SUBTYPE(pctx->btype) == TYPE_CMP_MAX) {
 			DEBUG_STAT_EN(double strt, en);
 
 			DEBUG_STAT_EN(strt = get_wtime_millis());
diff --git a/ppmd_compress.c b/ppmd_compress.c
index 3552a45..fafc362 100644
--- a/ppmd_compress.c
+++ b/ppmd_compress.c
@@ -114,8 +114,9 @@ ppmd_compress(void *src, uint64_t srclen, void *dst,
 	CPpmd8 *_ppmd = (CPpmd8 *)data;
 	uchar_t *_src = (uchar_t *)src;
 
-	if (btype & TYPE_COMPRESSED)
+	if (PC_TYPE(btype) == TYPE_COMPRESSED)
 		return (-1);
+
 	Ppmd8_RangeEnc_Init(_ppmd);
 	Ppmd8_Init(_ppmd, _ppmd->Order, PPMD8_RESTORE_METHOD_RESTART);
 	_ppmd->buf = (Byte *)dst;
diff --git a/utils/phash/extensions.h b/utils/phash/extensions.h
index b0fc4c0..c4a5b49 100644
--- a/utils/phash/extensions.h
+++ b/utils/phash/extensions.h
@@ -99,8 +99,8 @@ struct ext_entry {
 	{"xpi"	, TYPE_BINARY|TYPE_EXE, 3},
 	{"off"	, TYPE_BINARY|TYPE_EXE, 3},
 	{"pdf"	, TYPE_BINARY, 3},
-	{"jpg"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_JPEG, 3},
-	{"jpeg"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_JPEG, 4},
+	{"jpg"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX, 3},
+	{"jpeg"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX, 4},
 	{"png"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_GZ, 3},
 	{"mp3"	, TYPE_BINARY|TYPE_COMPRESSED, 3},
 	{"wma"	, TYPE_BINARY|TYPE_COMPRESSED, 3},
@@ -114,7 +114,8 @@ struct ext_entry {
 	{"flac"	, TYPE_BINARY|TYPE_COMPRESSED, 4},
 	{"pac"	, TYPE_BINARY|TYPE_COMPRESSED, 3},
 	{"gif"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_LZW, 3},
-	{"jp2"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_JPEG, 3},
+	{"jp2"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX, 3},
+	{"pjg"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX, 3},
 	{"gz"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_GZ, 2},
 	{"tgz"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_GZ, 3},
 	{"bz2"	, TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_BZ2, 3},
diff --git a/utils/phash/extensions.txt b/utils/phash/extensions.txt
index 0f58bd1..a349510 100644
--- a/utils/phash/extensions.txt
+++ b/utils/phash/extensions.txt
@@ -89,8 +89,8 @@ com,TYPE_BINARY|TYPE_EXE
 xpi,TYPE_BINARY|TYPE_EXE
 off,TYPE_BINARY|TYPE_EXE
 pdf,TYPE_BINARY
-jpg,TYPE_BINARY|TYPE_COMPRESSED|TYPE_JPEG
-jpeg,TYPE_BINARY|TYPE_COMPRESSED|TYPE_JPEG
+jpg,TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX
+jpeg,TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX
 png,TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_GZ
 mp3,TYPE_BINARY|TYPE_COMPRESSED
 wma,TYPE_BINARY|TYPE_COMPRESSED
@@ -104,7 +104,8 @@ ofr,TYPE_BINARY|TYPE_COMPRESSED
 flac,TYPE_BINARY|TYPE_COMPRESSED
 pac,TYPE_BINARY|TYPE_COMPRESSED
 gif,TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_LZW
-jp2,TYPE_BINARY|TYPE_COMPRESSED|TYPE_JPEG
+jp2,TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX
+pjg,TYPE_BINARY|TYPE_COMPRESSED|TYPE_CMP_MAX
 gz,TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_GZ
 tgz,TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_GZ
 bz2,TYPE_BINARY|TYPE_COMPRESSED|TYPE_COMPRESSED_BZ2
diff --git a/utils/phash/phash.c b/utils/phash/phash.c
index 5e2f340..b235b11 100644
--- a/utils/phash/phash.c
+++ b/utils/phash/phash.c
@@ -13,12 +13,12 @@
 /* small adjustments to _a_ to make values distinct */
 ub1 tab[] = {
 125,0,0,82,113,0,125,85,113,0,0,7,0,0,125,0,
-0,0,7,87,0,0,82,0,0,88,0,7,0,85,125,85,
-0,113,0,0,85,0,0,113,0,113,124,125,0,125,0,0,
-113,0,11,113,125,0,0,0,0,85,113,85,22,0,0,125,
-0,113,0,0,113,0,82,0,125,111,87,88,69,125,113,0,
-124,0,7,22,113,22,0,235,0,120,120,125,113,0,74,120,
-0,124,87,7,0,127,0,0,11,85,85,146,115,11,183,146,
+0,0,7,87,0,0,82,0,0,125,0,7,0,85,125,85,
+0,113,0,0,85,0,0,113,0,113,124,0,0,125,0,0,
+113,0,11,113,125,7,0,0,0,85,113,85,22,0,0,125,
+0,113,0,0,113,0,82,0,125,111,87,124,69,183,113,0,
+124,0,7,22,120,22,0,127,0,120,51,125,0,0,74,120,
+0,124,87,113,0,127,0,0,11,85,85,146,69,11,183,146,
 0,0,88,0,0,85,42,0,171,0,0,0,0,83,0,0,
 };
 
diff --git a/utils/phash/phash.h b/utils/phash/phash.h
index 74bd726..e422e47 100644
--- a/utils/phash/phash.h
+++ b/utils/phash/phash.h
@@ -8,7 +8,7 @@
 
 extern ub1 tab[];
 #define PHASHLEN 0x80  /* length of hash mapping table */
-#define PHASHNKEYS 133  /* How many keys were hashed */
+#define PHASHNKEYS 134  /* How many keys were hashed */
 #define PHASHRANGE 256  /* Range any input might map to */
 #define PHASHSALT 0x9e3779b9 /* internal, initialize normal hash */
 
diff --git a/utils/utils.h b/utils/utils.h
index 3a92355..464581d 100644
--- a/utils/utils.h
+++ b/utils/utils.h
@@ -232,31 +232,53 @@ struct fn_list {
  * Enumerated type constants for file type identification in pc_archive.
  */
 typedef enum {
+	/*
+	 * Primary Types.
+	 */
 	TYPE_UNKNOWN = 0,
 	TYPE_TEXT = 1,
 	TYPE_BINARY = 2,
 	TYPE_COMPRESSED = 4,
+	/*
+	 * Sub-types.
+	 */
 	TYPE_EXE = 8,
-	TYPE_JPEG = 12,
-	TYPE_MARKUP = 16,
-	TYPE_COMPRESSED_GZ = 20,
-	TYPE_COMPRESSED_LZW = 24,
-	TYPE_COMPRESSED_BZ2 = 28,
-	TYPE_COMPRESSED_ZIP = 32,
-	TYPE_COMPRESSED_ARJ = 36,
-	TYPE_COMPRESSED_ARC = 40,
-	TYPE_COMPRESSED_LH = 44,
-	TYPE_COMPRESSED_LZMA = 48,
-	TYPE_COMPRESSED_LZO = 52,
-	TYPE_COMPRESSED_UHARC = 56,
-	TYPE_COMPRESSED_ALZ = 60,
-	TYPE_COMPRESSED_ACE = 64,
-	TYPE_COMPRESSED_RAR = 68,
-	TYPE_COMPRESSED_LZ = 72,
-	TYPE_COMPRESSED_PPMD = 76,
-	TYPE_COMPRESSED_ZPAQ = 80
+	TYPE_CMP_MAX = 16,
+	TYPE_MARKUP = 24,
+	TYPE_COMPRESSED_GZ = 32,
+	TYPE_COMPRESSED_LZW = 40,
+	TYPE_COMPRESSED_BZ2 = 48,
+	TYPE_COMPRESSED_ZIP = 56,
+	TYPE_COMPRESSED_ARJ = 64,
+	TYPE_COMPRESSED_ARC = 72,
+	TYPE_COMPRESSED_LH = 80,
+	TYPE_COMPRESSED_LZMA = 88,
+	TYPE_COMPRESSED_LZO = 96,
+	TYPE_COMPRESSED_UHARC = 104,
+	TYPE_COMPRESSED_ALZ = 112,
+	TYPE_COMPRESSED_ACE = 120,
+	TYPE_COMPRESSED_RAR = 128,
+	TYPE_COMPRESSED_LZ = 136,
+	TYPE_COMPRESSED_PPMD = 144,
+	TYPE_COMPRESSED_ZPAQ = 152
 } data_type_t;
 
+/*
+ * Type identifier is an int with the following format.
+ *
+ *                Sub Type            Primary Type
+ *            (Numeric Value)    (Bit Positions - Flags)
+ *      _____________|_____________     ___|___
+ *     |                           |   |       |
+ *     .---------------------------------------.
+ *     |   |   |   |   |   |   |   |   |   |   |
+ * Bit 10                                         Bit 0
+ */
+#define	PC_TYPE_MASK	0x7
+#define	PC_SUBTYPE_MASK	0x7f8
+#define	PC_SUBTYPE(x)	((x) & PC_SUBTYPE_MASK)
+#define	PC_TYPE(x)	((x) & PC_TYPE_MASK)
+
 #ifndef _IN_UTILS_
 extern processor_info_t proc_info;
 #endif
diff --git a/zlib_compress.c b/zlib_compress.c
index dfd7ced..fddbc07 100644
--- a/zlib_compress.c
+++ b/zlib_compress.c
@@ -152,6 +152,19 @@ zlib_compress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
 	uchar_t *src1 = (uchar_t *)src;
 	z_stream *zs = (z_stream *)data;
 
+	/*
+	 * If the data is known to be compressed then certain types less compressed data
+	 * can be attempted to be compressed again for a possible gain. For others it is
+	 * a waste of time.
+	 */
+	if (PC_TYPE(btype) == TYPE_COMPRESSED && level < 7) {
+		int subtype = PC_SUBTYPE(btype);
+
+		if (subtype != TYPE_COMPRESSED_LZW &&
+		    subtype != TYPE_COMPRESSED_LZ && subtype != TYPE_COMPRESSED_LZO) {
+			return (-1);
+		}
+	}
 	ending = 0;
 	while (_srclen > 0) {
 		if (_srclen > SINGLE_CALL_MAX) {