Harry Potter schrieb am Freitag, 27. März 2020 um 01:04:32 UTC+1:

On Thursday, March 26, 2020 at 5:35:49 PM UTC-4, Harry Potter wrote:

I like your idea but admit that I don't fully understand it. :( Can you post some P-code to describe the technique? :)

Never mind: I get it. :) I'm applying your ideas now, but I'm running into some problems with your technique. :(

What's your progress so far?

On encode.su I wrote some thoughts about it and where it might loose efficiency. Some improvement ideas are also there. The main benefits in my humble view might come with text compression, as there are a lot of reused strings, where removing the length encoding for all of them could offset the need to have pointers into the decompressed block including for the first occurence of the string.

M.

I like the sound of your technique, but I don't fully understand how it works.  How do I specify where to insert an offset block? 

On Wednesday, February 12, 2020 at 11:35:36 AM UTC-5, Gerald R. Tamayo wrote:

Reduced Length LZ (RLLZ)

Let me share here my one way to output LZ77 codes, which may improve LZ77/LZSS or LZW.

LZ77:

LZ77 coding transmits <offset, length> codes. Many times in the compression process, the same matching string is encoded with an <offset> and the same <length> code. We can avoid transmitting the <length> code for many strings as well as not outputting a bit to identify a literal (as in LZSS) by the following:

Instead, read the whole input or block and gather duplicated or similar strings and encode <the whole string> (e.g., <length of the string (n>=2)> plus the actual string of characters) only this one time, <the number of the same matching strings (i.e., number of following offset codes)>, and its succeeding occurrences in the input block by transmitting only the said <offset> codes. (Or you can use an escape code to end the offset codes, perhaps BLOCK_SIZE.) Do this for all duplicated strings. This is actually a "Reduced Length LZ (RLLZ)".

The literals are outputted last *without bit flags* since they fall into the block buffers not covered or "not activated" by encoded strings. So just one array of literals can be outputted maybe at the end of the file, or, since block-based coding is more practical for shorter offset codes, at the end of the appropriately-sized block of <offsets>.

During decoding, the strings are written first in the output or write buffer using the offsets, and the literals "fill in" the unwritten positions or "holes" in the write buffer. This makes very compact LZ.


***

No need to output bit flags for literals or the number of following consecutive literals in some algorithms.

Then the completely filled write buffer is written to file.

That is, e.g. after gathering all distinct repeated strings on a block,

1) no transmission of <length> code for the next occurrences of the same string (but, in the simplest way, you have to output the number of succeeding strings);

2) transmitting the literals last (in the output buffer) means no need to transmit *bit flags* for literals (and matches). This is the novel or surprising idea here: deferred literals output;

3) if you know (LZT) LZ-Tamayo (2008) algorithm where it was demonstrated complete exclusion of the <length> code, this might as well be "LZ-Tamayo2". Should improve LZ77/LZSS/LZW based compressors. Decoding is also straight-forward.

Sorry that only now after a decade of LZT (2008 ) i am releasing this. I stopped coding compression in 2010 or 2011. (Note: it seems there is already "LZT" before i named my algorithm. LZ-Tischer.)
***

I have similar ideas as "rep-codes" in 2000s. My compression ideas are from the late 90s when i became interested again in data compression.

The ones i call here "holes" they call "gaps", even in LZW. But the idea of deferring transmission of literals for an output buffer avoids bit flags for both literals and matches which is still used in some explanations of ROLZ. Other algorithms still need to output the number of following literals, or literal_length. Deferred, meaning this is not an "online" algorithm.

-- Gerald R. Tamayo

(reposted, edited from Sept. 2018)

Hi, again!  I just figured out how to use your technique: I missed the part where you stated how to write the compressed block pointers.  I get it now, and, well, your technique worked!  :)  I had to kill a few of my techniques, but it was well worth it: I'm doing much better than before.  Thank you!

I was wrong.  I found a bug in my software,and now, I'm doing worse than without your technique, and I couldn't get back the numbers.  :(

Perhaps this simple illustration will help:

Reduced Length LZ (RLLZ)
(A very compact LZ. Made possible by deferred literals output.)

aacbbdeaab <- input source
0123456789 <- index

Compress:

Encode all strings first:

(size of string, string, number of occurrences of strings), [positions in file or block]

aa: (2, aa, 2), [0, 7]
bb: (2, bb, 1), [3]

Encode literals last: [cdeb]

Decompress:

Decode all strings first;
Decode literals.
(The strings and literals are nicely in their correct order or sequences.)

Perhaps this simple illustration will help:


Reduced Length LZ (RLLZ)
(A very compact LZ. Made possible by deferred literals output.)

aacbbdeaabb   <- input source
012345678910 <- index

encode all strings first:

(size of string, string, number of occurrences of strings), [positions in file or block]

aa: (2, aa, 2), [0, 7]
bb: (2, bb, 2), [3, 9]

encode literals last: [cde]

Decompress:

Decode all strings first;
Decode literals.
(The strings and literals are nicely in their correct order or sequences.)

On Wednesday, March 17, 2021 at 5:45:28 AM UTC-4, Gerald R. Tamayo wrote:

Perhaps this simple illustration will help:


Reduced Length LZ (RLLZ)
(A very compact LZ. Made possible by deferred literals output.)

aacbbdeaabb <- input source
012345678910 <- index

encode all strings first:

(size of string, string, number of occurrences of strings), [positions in file or block]

aa: (2, aa, 2), [0, 7]
bb: (2, bb, 2), [3, 9]

encode literals last: [cde]

Decompress:

Decode all strings first;
Decode literals.
(The strings and literals are nicely in their correct order or sequences.)

I did that then some performance optimizations and am doing *worse* than without your technique.  Maybe I'm doing something wrong.  I need to try it again.  :)

Reduced Length LZ (RLLZ)

Let me share here my one way to output LZ77 codes, which may improve LZ77/LZSS or LZW.

LZ77:

LZ77 coding transmits <offset, length> codes. Many times in the compression process, the same matching string is encoded with an <offset> and the same <length> code. We can avoid transmitting the <length> code for many strings as well as not outputting a bit to identify a literal (as in LZSS) by the following:

Instead, read the whole input or block  and gather duplicated or similar strings and encode <the whole string> (e.g., <length of the string (n>=2)> plus the actual string of characters) only this one time, <the number of the same matching strings (i.e., number of following offset codes)>, and its succeeding occurrences in the input block by transmitting only the said <offset> codes. (Or you can use an escape code to end the offset codes, perhaps BLOCK_SIZE.) Do this for all duplicated strings. This is actually a "Reduced Length LZ (RLLZ)".

The literals are outputted last *without bit flags* since they fall into the block buffers not covered or "not activated" by encoded strings. So just one array of literals can be outputted maybe at the end of the file, or, since block-based coding is more practical for shorter offset codes, at the end of the appropriately-sized block of <offsets>.

During decoding, the strings are written first in the output or write buffer using the offsets, and the literals "fill in" the unwritten positions or "holes" in the write buffer. This makes very compact LZ.


                ***

No need to output bit flags for literals or the number of following consecutive literals in some algorithms.

Then the completely filled write buffer is written to file.

That is, e.g. after gathering all distinct repeated strings on a block,

1) no transmission of <length> code for the next occurrences of the same string (but, in the simplest way, you have to output the number of succeeding strings);

2) transmitting the literals last (in the output buffer) means no need to transmit *bit flags* for literals (and matches). This is the novel or surprising idea here: deferred literals output;

3) if you know (LZT) LZ-Tamayo (2008) algorithm where it was demonstrated complete exclusion of the <length> code, this might as well be "LZ-Tamayo2". Should improve LZ77/LZSS/LZW based compressors. Decoding is also straight-forward.

Sorry that only now after a decade of LZT (2008 ) i am releasing this. I stopped coding compression in 2010 or 2011. (Note: it seems there is already "LZT" before i named my algorithm. LZ-Tischer.)
***

I have similar ideas as "rep-codes" in 2000s. My compression ideas are from the late 90s when i became interested again in data compression.

The ones i call here "holes" they call "gaps", even in LZW. But the idea of deferring transmission of literals for an output buffer avoids bit flags for both literals and matches which is still used in some explanations of ROLZ. Other algorithms still need to output the number of following literals, or literal_length. Deferred, meaning this is not an "online" algorithm.

-- Gerald R. Tamayo

(reposted, edited from Sept. 2018)

There are no "size" of holes to transmit, there are just characters written to holes, whatever the sizes of those holes, 1 char, 2 chars, or n chars, the literals are just written one by one into the holes.

Decode Output buffer (one appropriately-sized block):

[STRING..STRING....STRING.STRING.....STRING.STRING]
(2 holes, 4 holes, 1 hole, 5 holes, 1 hole)

The literals "nicely fit" into the holes in the write buffer, and they are in their *correct positions* in the file!, like the strings. Simply put, we still preserve the order or sequence of literals and strings.

Edit: <BLOCK_SIZE-1>

<the number of the same matching strings (i.e., number of following offset codes)>, and its succeeding occurrences in the input block by transmitting only the said <offset> codes. (Or you can use an escape code to end the offset codes, perhaps <BLOCK_SIZE-1>.)

-- Gerald

I like your idea but admit that I don't fully understand it.  :(  Can you post some P-code to describe the technique?  :)

On Thursday, March 26, 2020 at 5:35:49 PM UTC-4, Harry Potter wrote:

I like your idea but admit that I don't fully understand it.  :(  Can you post some P-code to describe the technique?  :)

Never mind: I get it.  :)   I'm applying your ideas now, but I'm running into some problems with your technique.  :(

Do not immediately write the codes to output file. (That's what most LZ compressors do.) Buffer it. So you can output strings to the output buffer first, then the literals. Then write the whole buffer to output file. The original sequence or order of the literals is still preserved.

Do not immediately write the codes to output file. (That's what most LZ compressors do.) Buffer it. So you can output strings to the output buffer first, then the literals. Then write the whole buffer to output file. The original sequence or order of the literals and strings is still preserved.

Do not immediately write the codes to output file. (That's what most LZ compressors do.) Buffer it. So you can output strings to the output buffer first, then the literals. Then write the whole buffer to output file. The original sequence or order of the literals and strings is still preserved.

I did that then some performance optimizations and am doing *worse* than without your technique. Maybe I'm doing something wrong. I need to try it again. :)

Maybe the following improved example will help.

Here is simple illustration for RLLZ, to be complete:

Reduced Length LZ (RLLZ)
(A very compact LZ. Made possible by deferred literals output.)

aacbcdeaabc <- input source
012345678910 <- index

Compress:

Encode all duplicated strings first:
Very simple:
(<size of string>, <string>, <number of "next" occurrences of string>), [positions in file or block (no match lengths needed)]

aa: (2, aa, 1), [0, 7]
bc: (2, bc, 1), [3, 9]

Better:
aa: (2, aa, 1-1=0), [0, 7]
bc: (2, bc, 1-1=0), [3, 9]

Encode literals last: [cde]


Decompress:

Decode all strings first;
Read/Get literals. (The literals are then inserted into the holes. Whatever the size of those holes randomly positioned in the write buffer, they are filled by the literals.)

The strings and literals are nicely in their correct order or sequences. Write to actual file.

Maybe there's a problem with my code.  (?)  I pasted the relevant code here.  It uses Assembler, but I believe the original C code is still in the file.
--------------------------------
void CompBin01 (void)
{
//signed char c2; //Holds value for len. current
// int c2;
// unsigned /*j1,*/ j3, j4, j5;
// unsigned i, j, j2, k, m, n, o=0;
//vz.InPos=0;
k=0; vz.InPos=0;
for (; vz.InPos<vz.InEnd; )
{
//printc ('.');
printu (vz.InPos); printcr ();
/* if (Buffer2[vz.InPos>>3]&(1<<(vz.InPos&7))) {
//getkey ();
for (i=j=m=0; i<NumRLLZEnt; ++i) {
if (RLLZBuf[i].len>j && !memcmp(&InBuffer[vz.InPos], &InBuffer[RLLZBuf[i].loc], RLLZBuf[i].len)) {
j=RLLZBuf[i].len; m=i;
}
}
//if (j==0) {CompBin01U (); ++vz.InPos; continue;}
if (j==0) {WriteB (InBuffer[vz.InPos]); ++vz.InPos; continue;}
//CompBin01Ue ();
++backcol;
vz.InPos+=j;
} else {
//CompBin01U (); ++vz.InPos;
WriteB (InBuffer[vz.InPos]); ++vz.InPos;
}*/
//if (!(Buffer2[vz.InPos>>3]&(1<<(vz.InPos&7)))) WriteB (InBuffer[vz.InPos]);
//else ++backcol;
// if (!(Buffer2[vz.InPos>>3]&(1<<(vz.InPos&7)))) {
// WriteB (InBuffer[vz.InPos]);
// //++backcol;
// }
// ++vz.InPos;
if ((Buffer2[vz.InPos>>3]&(1<<(vz.InPos&7)))) {
//CompBin01Ue ();
writebuffer ();
for (; Buffer2[vz.InPos>>3]&(1<<(vz.InPos&7)); ++vz.InPos);
//++backcol;
} else {
CompBin01U ();
++vz.InPos;
}
}
writeoutlast ();
}
static unsigned p;

void GetRLLZBlocks (void)
{
unsigned i, j, k, l, m, n, o;
unsigned q;
struct RLLZBuf* r;
for (i=NumRLLZEnt=0; i<vz.InEnd; ) {
printu (i); printcr ();
if ((lenc=GetLZW (i))>=4) {
o=techLZW.pos-(unsigned)&InBuffer;
//if (GetLZW(i+1)>=lenc+1) {++i; continue;}
for (j=0; j<NumRLLZEnt; ++j) {
if (RLLZBuf[j].len==lenc &&
!memcmp (&InBuffer[o], &InBuffer[RLLZBuf[j].loc], RLLZBuf[j].len))
break;
}
if (j==NumRLLZEnt) {
if (NumRLLZEnt>=1000) {
//++backcol;
++i; continue;
}
RLLZBuf[j].loc=o;
RLLZBuf[j].len=lenc;
for (k=0; k<lenc; k++) {
//Buffer3[(o+k)>>3]|=1<<(o+k&7);
}
RLLZBuf[j].numoccur=1;
++NumRLLZEnt;
}
++RLLZBuf[j].numoccur;
for (k=0; k<lenc; k++) {
//++backcol;
//Buffer3[(i+k)>>3]|=1<<(i+k&7);
}
i+=techLZW.size;
} else {
a: ++i;
}
}
for (i=0; i<NumRLLZEnt; ++i) {
//++backcol;
if (RLLZBuf[i].numoccur<5-(RLLZBuf[i].len>=5?2:0)) {
for (j=i; j<NumRLLZEnt; ++j) {
memcpy (&RLLZBuf[j], &RLLZBuf[j+1], sizeof (RLLZBuf[0]));
} --NumRLLZEnt; --i;
}
}
prints ("rllz="); printu (NumRLLZEnt); printcr (); getkey ();
WriteDist (NumRLLZEnt, 1024);
for (o=0; o<NumRLLZEnt; ++o) {
i=2;
lenc=RLLZBuf[o].len;
if (lenc<6) {
WriteDist (lenc-2+1-i, 6-i);
} else if (lenc< 11) {
WriteDist (0, 6-i);
//writeoutf_10 ();
WriteDist (lenc-6,  11-6);
} else if (lenc<18) {
WriteDist (5-i, 6-i);
//writeoutf_10 ();
WriteDist ((lenc- 11), (22- 11));
} else {
WriteDist (5-i, 6-i);
WriteDist ((18- 11)+(lenc-18)%4, (22- 11));
//writeoutf_01 ();
//WriteDist (11-7+((lenc-11)&5), 17-7);
WriteDist ((lenc-18)/4, (38-18+3)/4);
//WriteDist ((lenc-18), (57-18));
}
for (j=0; j<lenc; ++j) {
WriteB (InBuffer[RLLZBuf[o].loc+j]);
} k=0; m=-1;

for (l=n=0; l<vz.InEnd; )
{
//++backcol;
lenc=0; p=-1;
//if (Buffer2[vz.InPos>>3]&(1<<(vz.InPos&7))) {
//if (GetLZW(l)>=4) {
//if (!(Buffer3[l]>>(l>>3)&1<<(i&7))) {++l; continue;}
//if (InBuffer[l]!=InBuffer)
//r=&RLLZBuf[0];
cin=&InBuffer[l];
/*for (j=0; j<NumRLLZEnt; ++j) {

if ((q=r->len)>lenc &&
*cin==InBuffer[i=r->loc] &&
//(Buffer3[l>>3]&1<<(l&7)) &&
memcmp2(&InBuffer[i], q))
{lenc=q; p=j;}
++r;
}*/
__asm__ (
"\tlda\t#0\n"
"\tsta\t_i\n"
//"\tlda\tNumRLLZEnt+1\n"
"\tsta\t_i+1\n"
"\tlda\t#<_RLLZBuf\n"
"\tsta\tptr2\n"
"\tlda\t#>_RLLZBuf\n"
"\tsta\tptr2+1\n"

// "\tlda\t_cin\n"
// "\tsta\tptr4\n"
// "\tlda\t_cin+1\n"
// "\tsta\tptr4+1\n"

"@lp1:\n"
"\tldy\t#2\n"
"\tlda\t_lenc\n"
"\tcmp\t(ptr2),y\n"
"\tbeq\t@skip\n"
"\tbcs\t@skip\n"
"\tlda\t#<_InBuffer\n"
"\tldy\t#0\n"
"\tclc\n"
"\tadc\t(ptr2),y\n"
"\tsta\tptr3\n"
"\tiny\n"
"\tlda\t#>_InBuffer\n"
"\tadc\t(ptr2),y\n"
"\tsta\tptr3+1\n"

"\tldy\t#0\n"
"\tlda\t(_cin),y\n"
"\tcmp\t(ptr3),y\n"
"\tbne\t@skip\n"

"\tldy\t#2\n"
"\tlda\t(ptr2),y\n"
"\ttax\n"
"\ttay\n"
"\tdey\n"
"@lp2:\n"
"\tlda\t(_cin),y\n"
"\tcmp\t(ptr3),y\n"
"\tbne\t@skip\n"
"\tdey\n"
"\tbne\t@lp2\n"

"\tlda\t_i\n"
"\tsta\t_p\n"
"\tlda\t_i+1\n"
"\tsta\t_p+1\n"
"\tstx\t_lenc\n"

"@skip:\n"
"\tlda\tptr2\n"
"\tclc\n"
"\tadc\t#8\n"
"\tsta\tptr2\n"
"\tbcc\t@skip2\n"
"\tinc\tptr2+1\n"
"@skip2:\n"
"\tinc\t_i\n"
"\tbne\t@skip3\n"
"\tinc\t_i+1\n"
"@skip3:\n"
"\tlda\t_i+1\n"
"\tcmp\t_NumRLLZEnt+1\n"
"\tbcc\t@lp1\n"
"\tlda\t_i\n"
"\tcmp\t_NumRLLZEnt\n"
"\tbcc\t@lp1\n"
);
if (lenc && p==o) {
//++backcol;
for (i=0; i<lenc; i++) {
Buffer2[(l+i)>>3]|=1<<(l+i&7);
//++backcol;
}
WriteDist (l-k, vz.InEnd-k+1);
//k=l; l+=lenc;
k=l+lenc; l+=lenc;
} else {
++l;
}
} WriteDist (vz.InEnd-k, vz.InEnd+1-k);
printu (o); printcr ();
}
//printf (">>> %X\n", &Buffer2); getkey ();
}
--------------------------------
WriteDist() writes the information using as few bits as possible.

Maybe the following improved example will help.

Here is simple illustration for RLLZ, to be complete:

Reduced Length LZ (RLLZ)
(A very compact LZ. Made possible by deferred literals output.)

aacbcdeaabc <- input source
012345678910 <- index

Compress:

Encode all duplicated strings first:
Very simple:
(<size of string>, <string>, <number of "next" occurrences of string>), [positions in file or block (no match lengths needed)],
   :

aa: (2, aa, 1), [0, 7]
bc: (2, bc, 1), [3, 9]

Better:
aa: (2, aa, 1-1=0), [0, 7]
bc: (2, bc, 1-1=0), [3, 9]

Encode end of string code stream: (0, , );

Encode literals last: [cde].


Decompress:

Decode all strings first;
Read/Get literals. (The literals are then inserted into the holes. Whatever the size of those holes randomly positioned in the write buffer, they are filled by the literals.)

The strings and literals are nicely in their correct order or sequences. Write to actual file.

On Saturday, October 2, 2021 at 11:54:18 PM UTC+8, Gerald R. Tamayo wrote:

Maybe the following improved example will help.

Here is simple illustration for RLLZ, to be complete:
Reduced Length LZ (RLLZ)
(A very compact LZ. Made possible by deferred literals output.)
aacbcdeaabc <- input source
012345678910 <- index

Compress:

Encode all duplicated strings first:
Very simple:
(<size of string>, <string>, <number of "next" occurrences of string>), [positions in file or block (no match lengths needed)],
:

aa: (2, aa, 1), [0, 7]
bc: (2, bc, 1), [3, 9]

Better:
aa: (2, aa, 1-1=0), [0, 7]
bc: (2, bc, 1-1=0), [3, 9]

Encode end of string code stream: (0, , );

Encode literals last: [cde].
Decompress:

Decode all strings first;
Read/Get literals. (The literals are then inserted into the holes. Whatever the size of those holes randomly positioned in the write buffer, they are filled by the literals.)

The strings and literals are nicely in their correct order or sequences. Write to actual file.

"Deferred literals transmission" in Reduced Length LZ (#RLLZ) #algorithm actually has many practical uses in #data #compression! Not just in the way I described RLLZ.

"Deferred literals transmission" in Reduced Length LZ (#RLLZ) #algorithm actually has many practical uses in #data #compression! Not just in the way I described RLLZ.

You can use "deferred literals transmission" if you want. Permission granted, no strings attached. ;)

I want to implement this on small 8-bit and 16-bit platforms where I might not have enough memory to store a whole file.  Is there

I want to implement this on small 8-bit and 16-bit platforms where I might not have enough memory to store a whole file. Is there a way to get this to work only on a piece of a file?  Or maight I have to divide the file into smaller pieces?

BTW, sorry about the last, incomplete post.  I pressed the wrong key.  :(