zefie/minisrv
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'lzpf-compression' into dev

Browse Source
This commit is contained in:
zefie
2021-09-10 23:37:55 -04:00
parent 7d9a465bfb 6af01ce08f
commit c925cdf34e
1 changed files with 9 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
zefie_wtvp_minisrv/WTVLzpf.js
View File

@@ -40,11 +40,11 @@ class WTVLzpf {
* - Bytes that don't change the length of the bit stream: 024:MW_kvy
* - The rest will increase the length of bit stream
*
* I don't know what process they used to build this dictionary. I assume they
* I don't know what process they used to build this table. I assume they
* frequency-scanned a bunch of HTML files they had.
*
* Using ISO-8859-1 chracter encoding. Didn't seem like they used a different
* dictionary for Japan builds (ISO-2022-JP).
* Using Windows-1252 (based off of ISO-8859-1) chracter encoding to fill in this table. Didn't
* seem like they used a different table for Japan builds (ISO-2022-JP).
**/
nomatchEncode = [
/* [FLATTENED HUFFMAN CODE, CODE BIT LENGTH] */
@@ -138,13 +138,13 @@ class WTVLzpf {
/**
* This is the dictionary that reduces the size based on repeated patterns in the file.
* This is the table that reduces the size based on repeated patterns in the file.
*
* When we find a byte match in the ring buffer we use this dictionary to encode the length of the matched bytes.
* When we find a byte match in the ring buffer we use this table to encode the length of the matched bytes.
*
* - These are intentionally 32-bit. The leftmost flag bit is 1 in each of these to tell the decoder to use match decoding.
* - LZP hash bits are used to encode the position where the matched bytes start.
* - We're allowed to match up to 298 bytes before we can't encode more (we need an entry in this dictionary for each byte more).
* - We're allowed to match up to 298 bytes before we can't encode more (we need an entry in this table for each byte more).
* - We can reach for matches 65KB behind the current LZ cursor (65KB is the ring buffer size and highest a 16-bit hash can reach).
**/
matchEncode = [
@@ -422,7 +422,7 @@ class WTVLzpf {
}
if (this.ring_bufer_index == 0xFFFF) {
// We never seen this byte before so we encode it with our Huffman dictionary.
// We never seen this byte before so we encode it with our Huffman table.
code_length = this.nomatchEncode[byte][1];
code = this.nomatchEncode[byte][0] << 0x10;
} else if (byte == this.ring_buffer[this.ring_bufer_index] && compress_data) {
@@ -431,7 +431,7 @@ class WTVLzpf {
this.ring_bufer_index = (this.ring_bufer_index + 1) & 0x1FFF;
this.compression_mode = 4;
} else {
// We've seen these bytes before but the index in the ring buffer doesn't match so we revert to our neat Huffman dictionary
// We've seen these bytes before but the index in the ring buffer doesn't match so we revert to our neat Huffman table
// We add 1 flag bit of 0 to account for the fact we've had a hash table hit but no hit in the ring buffer.
code_length = this.nomatchEncode[byte][1] + 1;
code = this.nomatchEncode[byte][0] << 0x0F;
@@ -483,7 +483,7 @@ class WTVLzpf {
this.AddBits((this.checksum << 0x18) & 0xFFFFFFFF, 0x08);
// If we have leftover bits then add it.
if (this.current_bits != 0x00) {
if (this.current_bit_length > 0) {
this.AddByte((this.current_bits >>> 0x18) & 0xFF);
}