From f432fa306ad37a634aad7390cc08cb617e37e488 Mon Sep 17 00:00:00 2001
From: Pompolic <pompolic@special-circumstanc.es>
Date: Tue, 26 Oct 2021 20:57:59 +0200
Subject: [PATCH] EarlyChange support
---
lzw.c | 63 +++++++++++++++++++++++------------------------------------
lzw.h | 7 ++++---
pdf.c | 17 +++++++++++-----
3 files changed, 40 insertions(+), 47 deletions(-)
diff --git a/lzw.c b/lzw.c
index c80eefd..304c88d 100644
--- a/lzw.c
+++ b/lzw.c
@@ -1,7 +1,6 @@
/* Table for storing sequences represented by an LZW code */
// XXX lookup is O(1) like this, but maybe memory use will be bad
// XXX unify lzw_context_t and lzwspec
-// XXX EarlyChange support
#include <hammer/hammer.h>
#include <hammer/glue.h>
// malloc, free
@@ -10,30 +9,18 @@
#include <string.h>
#include "lzw.h"
-//XXX lzw.h
-/*
-typedef struct LZW_context_S
-{
- *
- * Table for storing sequences represented by an LZW code
- * 0-255, and 256 are special, representing literals, and the reset code. We could explicitly pre-fill them, but it's probably not necessary.
- *
- HBytes * lzw_code_table[4096];
-
- *
- * Holds the next expected LZW code. We also use this for telling LZW_9bitcodeword, LZW_10bitcodeword, etc. apart. Parses fail if "next" is larger than what can be represented on that many bits.
- *
- int next;
-
- *
- * Previous LZW code, used to construct the next string added to the table.
- *
- uint64_t old;
-} LZW_context_T;
-*/
FILE *debug; // DEBUG
+#define BITLIMIT_9 (ctx->earlychange ? 511 : 512)
+#define BITLIMIT_10 (ctx->earlychange ? 1023 : 1024)
+#define BITLIMIT_11 (ctx->earlychange ? 2047 : 2048)
+/*
+ * Since bit lengths larger than 12 aren't allowed, EarlyChange doesn't matter here.
+ */
+#define BITLIMIT_12 4096
+
+
void LZW_clear_table(LZW_context_T *ctx)
{
/*
@@ -113,9 +100,9 @@ validate_LZW_9bitcodeword(HParseResult *p, void *u)
LZW_context_T * ctx = (LZW_context_T *) u;
//fprintf(debug, "9 bit code: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
- if (ctx->next < 511) // DEBUG
+ if (ctx->next < BITLIMIT_9) // DEBUG
assert(H_CAST_UINT(p->ast) <= ctx->next); // DEBUG
- return (ctx->next < 511); // XXX: parameterize codeword boundaries via EarlyChange
+ return (ctx->next < BITLIMIT_9);
}
bool
@@ -124,9 +111,9 @@ validate_LZW_10bitcodeword(HParseResult *p, void *u)
LZW_context_T * ctx = (LZW_context_T *) u;
//fprintf(debug, "10 bit code: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
- if (ctx->next >= 511 && ctx->next < 1023) // DEBUG
+ if (ctx->next >= BITLIMIT_9 && ctx->next < BITLIMIT_10) // DEBUG
assert(H_CAST_UINT(p->ast) <= ctx->next); // DEBUG
- return (ctx->next >= 511 && ctx->next < 1023);
+ return (ctx->next >= BITLIMIT_9 && ctx->next < BITLIMIT_10);
}
bool
@@ -135,9 +122,9 @@ validate_LZW_11bitcodeword(HParseResult *p, void *u)
LZW_context_T * ctx = (LZW_context_T *) u;
//fprintf(debug, "11 bit code: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
- if (ctx->next >= 1023 && ctx->next < 2047) // DEBUG
+ if (ctx->next >= BITLIMIT_10 && ctx->next < BITLIMIT_11) // DEBUG
assert(H_CAST_UINT(p->ast) <= ctx->next); // DEBUG
- return (ctx->next >= 1023 && ctx->next < 2047);
+ return (ctx->next >= BITLIMIT_10 && ctx->next < BITLIMIT_11);
}
bool
@@ -146,9 +133,9 @@ validate_LZW_12bitcodeword(HParseResult *p, void *u)
LZW_context_T * ctx = (LZW_context_T *) u;
//fprintf(debug, "12 bit code: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
- if (ctx->next >= 2047 && ctx->next < 4095) // DEBUG
+ if (ctx->next >= BITLIMIT_11 && ctx->next < BITLIMIT_12) // DEBUG
assert(H_CAST_UINT(p->ast) <= ctx->next); // DEBUG
- return (ctx->next >= 2047 && ctx->next < 4095);
+ return (ctx->next >= BITLIMIT_11 && ctx->next < BITLIMIT_12);
}
bool
@@ -158,7 +145,7 @@ validate_LZW_9bitlitspec(HParseResult *p, void *u)
//fprintf(debug, "9 bit lit: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
uint64_t code = H_CAST_UINT(p->ast);
- return (ctx->next < 511 && code < 258);
+ return (ctx->next < BITLIMIT_9 && code < 258);
}
bool
@@ -168,7 +155,7 @@ validate_LZW_10bitlitspec(HParseResult *p, void *u)
//fprintf(debug, "10 bit lit: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
uint64_t code = H_CAST_UINT(p->ast);
- return (ctx->next >= 511 && ctx->next < 1023 && code < 258);
+ return (ctx->next >= BITLIMIT_9 && ctx->next < BITLIMIT_10 && code < 258);
}
bool
@@ -178,7 +165,7 @@ validate_LZW_11bitlitspec(HParseResult *p, void *u)
//fprintf(debug, "11 bit lit: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
uint64_t code = H_CAST_UINT(p->ast);
- return (ctx->next >= 1023 && ctx->next < 2047 && code < 258);
+ return (ctx->next >= BITLIMIT_10 && ctx->next < BITLIMIT_11 && code < 258);
}
bool
@@ -188,7 +175,7 @@ validate_LZW_12bitlitspec(HParseResult *p, void *u)
//fprintf(debug, "12 bit lit: %lu, next: %u\n", p->ast->uint, ctx->next); // DEBUG
//fflush(debug); // DEBUG
uint64_t code = H_CAST_UINT(p->ast);
- return (ctx->next >= 2047 && ctx->next < 4095 && code < 258);
+ return (ctx->next >= BITLIMIT_11 && ctx->next < BITLIMIT_12 && code < 258);
}
bool
@@ -429,6 +416,7 @@ void init_LZW_parser()
lit->len = 1;
context->lzw_code_table[i] = lit;
}
+ context->earlychange = 1;
H_VDRULE(LZW_9bitcodeword, h_bits(9, false), context);
H_VDRULE(LZW_10bitcodeword, h_bits(10, false), context);
@@ -466,12 +454,8 @@ HParseResult* parse_LZW_data(const uint8_t* input, size_t length)
return res;
}
-void set_LZW_context(LZW_context_T *ctx)
-{
- *context = *ctx; // XXX unnecessary, just clear context before each parse
-}
-void clear_LZW_context()
+void init_LZW_context(int earlychange)
{
for(int i = 258; i < 4096; ++i)
{
@@ -484,4 +468,5 @@ void clear_LZW_context()
}
context->next = 258;
context->old = 257; //XXX: guaranteed to segfault if old isn't set before
+ context->earlychange = earlychange;
}
diff --git a/lzw.h b/lzw.h
index d4f4308..2525004 100644
--- a/lzw.h
+++ b/lzw.h
@@ -25,14 +25,15 @@ typedef struct LZW_context_S
/*
* EarlyChange = 1 means the bit size is increased "one code early" (Early change = 0 is "code length increases shall be postponed as long as possible"
*/
- //int earlychange;
+ int earlychange;
} LZW_context_T;
HParser * p_lzwdata; // XXX can be internal
void init_LZW_parser();
HParseResult * parse_LZW_data(const uint8_t* input, size_t length);
-void set_LZW_context(LZW_context_T *ctx);
-void clear_LZW_context();
+//void set_LZW_context(LZW_context_T *ctx);
+//void clear_LZW_context();
+void init_LZW_context(int earlychange);
#endif // PDF_LZW_H
diff --git a/pdf.c b/pdf.c
index 8e38bbb..976433a 100644
--- a/pdf.c
+++ b/pdf.c
@@ -3337,6 +3337,7 @@ LZWDecode(const Dict *parms, HBytes b, HParser *p)
int done;
//int ret;
const HParsedToken *v;
+ int earlychange;
/* set up the predictor (if any) */
#define SETPARM(VAR,STR) do { \
@@ -3385,6 +3386,16 @@ LZWDecode(const Dict *parms, HBytes b, HParser *p)
err(1, "LZWDecode");
}
+ v = dictentry(parms, "EarlyChange");
+ if(v != NULL && v->token_type == TT_SINT && v->sint == 0)
+ {
+ earlychange = 0;
+ }
+ else
+ {
+ earlychange = 1;
+ }
+
//lzwspec *lzw_spec = new_lzw_spec(&b);
//bind_lzw_spec(lzw_spec);
@@ -3395,12 +3406,8 @@ LZWDecode(const Dict *parms, HBytes b, HParser *p)
//}
//done = depredict(&pred, cur_lzw_spec->lzw_buf, cur_lzw_spec->write_head-1);
//assert(!done); // XXX ITERATIVE
- LZW_context_T * ctx = malloc(sizeof(LZW_context_T));
- ctx->next = 258;
- clear_LZW_context();
+ init_LZW_context(earlychange);
tmp_res = parse_LZW_data(b.token, b.len);
- //clear_LZW_context();
- free(ctx);
if(!tmp_res)
{
--
GitLab