Heads up, I might have actually squashed a few bugs during the time you were modifying it. You might want to change your parts of code that I have edited.
2 Likes
chatgpt will turn the code into js
1 Like
defining all functions as variables will have no improvement in fact I believe it could slow it down more, I tried to test it myself though and I couldn’t see any benefits when using locally declared functions.
this used to be a problem of the past though I remember it being discussed
try it for yourself replace the insert with table.insert if you want to try
local insert = table.insert
local Table1 = {}
local start1 = tick()
for i = 1, 10000000 do
insert(Table1,i)
end
print("code took "..tick() - start1.." ms!")
task.wait()
local start2 = tick()
local Table2 = {}
for i = 1, 10000000 do
table.insert(Table2,i)
end
print("code took "..tick() - start2.." ms!")
Most of the time the second print will show that it took slightly more but it doesnt have to do with the speed of the function but the first table slowing the code down switching the functions will reveal that again the second print will be slower
1 Like
I consider comparing loops with a compression algorithm invalid, try it with the actual module (and its versions) with multiple iterations.
1 Like
unfortunately, the revised module still seems to choke on bytes \0 and \255. this seems to be caused by off-by-one errors on lines 27 and 39 (1, 31 and 1, 255 - 181; should seemingly be 0, 31 and 1, 255 - 180).
however, my attempted fix causes a different error to appear, on line 76 due to tobase92 being called with a nil argument.
i don’t understand the code enough to properly fix this. can somebody more knowledgeable please take a look?
2 Likes
I actually fixed it a while back for myself, but I never bothered to post a new reply, so here you go.
-- Module by 1waffle1 and boatbomber, optimized and fixed by iiau
-- https://devforum.roblox.com/t/text-compression/163637/37
local dictionary = {} -- key to len
do -- populate dictionary
local length = 0
for i = 32, 127 do
if i ~= 34 and i ~= 92 then
local c = string.char(i)
dictionary[c] = length
dictionary[length] = c
length = length + 1
end
end
end
local escapemap_126, escapemap_127 = {}, {}
local unescapemap_126, unescapemap_127 = {}, {}
local blacklisted_126 = { 34, 92 }
for i = 126, 180 do
table.insert(blacklisted_126, i)
end
do -- Populate escape map
-- represents the numbers 0-31, 34, 92, 126, 127 (36 characters)
-- and 128-180 (52 characters)
-- https://devforum.roblox.com/t/text-compression/163637/5
for i = 0, 31 + #blacklisted_126 do
local b = blacklisted_126[i - 31]
local s = i + 32
-- Note: 126 and 127 are magic numbers
local c = string.char(b or i)
local e = string.char(s + (s >= 34 and 1 or 0) + (s >= 91 and 1 or 0))
escapemap_126[c] = e
unescapemap_126[e] = c
end
for i = 1, 255 - 180 do
local c = string.char(i + 180)
local s = i + 34
local e = string.char(s + (s >= 92 and 1 or 0))
escapemap_127[c] = e
unescapemap_127[e] = c
end
end
local function escape(s)
-- escape the control characters 0-31, double quote 34, backslash 92, Tilde 126, and DEL 127 (36 chars)
-- escape characters 128-180 (53 chars)
return string.gsub(string.gsub(s, '[%c"\\\126-\180]', function(c)
return "\126" .. escapemap_126[c]
end), '[\181-\255]', function(c)
return "\127" .. escapemap_127[c]
end)
end
local function unescape(s)
return string.gsub(string.gsub(s, "\127(.)", function(e)
return unescapemap_127[e]
end), "\126(.)", function(e)
return unescapemap_126[e]
end)
end
local b93Cache = {}
local function tobase93(n)
local value = b93Cache[n]
if value then
return value
end
local c = n
value = ""
repeat
local remainder = n % 93
value = dictionary[remainder] .. value
n = (n - remainder) / 93
until n == 0
b93Cache[c] = value
return value
end
local b10Cache = {}
local function tobase10(value)
local n = b10Cache[value]
if n then
return n
end
n = 0
for i = 1, #value do
n = n + math.pow(93, i - 1) * dictionary[string.sub(value, -i, -i)]
end
b10Cache[value] = n
return n
end
local function compress(text)
assert(type(text) == "string", "bad argument #1 to 'compress' (string expected, got " .. typeof(text) .. ")")
local dictionaryCopy = table.clone(dictionary)
local key, sequence, size = "", {}, #dictionary
local width, spans, span = 1, {}, 0
local function listkey(k)
local value = tobase93(dictionaryCopy[k])
local valueLength = #value
if valueLength > width then
width, span, spans[width] = valueLength, 0, span
end
table.insert(sequence, string.rep(" ", width - valueLength) .. value)
span += 1
end
text = escape(text)
for i = 1, #text do
local c = string.sub(text, i, i)
local new = key .. c
if dictionaryCopy[new] then
key = new
else
listkey(key)
key = c
size += 1
dictionaryCopy[new] = size
dictionaryCopy[size] = new
end
end
listkey(key)
spans[width] = span
return table.concat(spans, ",") .. "|" .. table.concat(sequence)
end
local function decompress(text)
assert(type(text) == "string", "bad argument #1 to 'decompress' (string expected, got " .. typeof(text) .. ")")
local dictionaryCopy = table.clone(dictionary)
local sequence, spans, content = {}, string.match(text, "(.-)|(.*)")
local groups, start = {}, 1
for span in string.gmatch(spans, "%d+") do
local width = #groups + 1
groups[width] = string.sub(content, start, start + span * width - 1)
start = start + span * width
end
local previous
for width, group in ipairs(groups) do
for value in string.gmatch(group, string.rep(".", width)) do
local entry = dictionaryCopy[tobase10(value)]
if previous then
if entry then
table.insert(dictionaryCopy, previous .. string.sub(entry, 1, 1))
else
entry = previous .. string.sub(previous, 1, 1)
table.insert(dictionaryCopy, entry)
end
table.insert(sequence, entry)
else
sequence[1] = entry
end
previous = entry
end
end
return unescape(table.concat(sequence))
end
return { compress = compress, decompress = decompress }
It turns out I didn’t need to change it to base-92. Instead I found out that the escapemap had overlap with character codes required to unescape it. So I have to make an entire dictionary just for unescaping the escape characters.
I’m currently using this live production in one of my games, Felandia, serving hundreds of daily players and I have not experienced any issues whatsoever.
Enjoy!
8 Likes
I’m exporting training data for a neural network and this compressed my json file from 31 script objects (199999 characters per script) to just 13. Good stuff!
1 Like
I’ve been rewriting this code into Typescript yesterday for my Datastore optimizations and sharding mechanics.
However as I read conversation, it looks like there are many improved versions, would someone point out which of these versions are production stable?
Thanks alot!
2 Likes
I went through most (some had problems), but the one modified by @iiau has been working great.
2 Likes
[EDIT] By using maybeyield from this reply by Anaminus, post Reply from Anaminus I was able to yield every .5 sec or so to avoid exhaust from occouring, just leaving this if anyone else run into this issue
Great module! One issue I run into is when compressing big texts is the “Script timeout: exhausted allowed execution time” Would it be possible to implement some way where you can customize how much of the processing power the system uses, lets say 70% of cpu or something along that?
1 Like
Hello. I have been using this module for around 2 years now to compress image data for CanvasDraw, which is my pixel-based graphics library for roblox.
I am wondering if there’s any room to further improve performance as I am having lag spike issues compressing a large 20,000 to 130,000 character strings.
I am using the optimised version edited by boatbomber:
--!native
-- Module by 1waffle1, optimized by boatbomber
-- https://devforum.roblox.com/t/text-compression/163637
local gsub = string.gsub
local sub = string.sub
local insert = table.insert
local rep = string.rep
local dictionary = {}
do -- populate dictionary
local length = 0
for i = 32, 127 do
if i ~= 34 and i ~= 92 then
local c = string.char(i)
dictionary[c], dictionary[length] = length, c
length = length + 1
end
end
end
local escapemap = {}
do -- Populate escape map
for i = 1, 34 do
i = ({ 34, 92, 127 })[i - 31] or i
local c, e = string.char(i), string.char(i + 31)
escapemap[c], escapemap[e] = e, c
end
end
local function escape(s)
return gsub(s, '[%c"\\]', function(c)
return "\127" .. escapemap[c]
end)
end
local function unescape(s)
return gsub(s, "\127(.)", function(c)
return escapemap[c]
end)
end
local function copy(t)
local new = {}
for k, v in pairs(t) do
new[k] = v
end
return new
end
local b93Cache = {}
local function tobase93(n)
local value = b93Cache[n]
if value then
return value
end
value = ""
repeat
local remainder = n % 93
value = dictionary[remainder] .. value
n = (n - remainder) / 93
until n == 0
b93Cache[n] = value
return value
end
local b10Cache = {}
local function tobase10(value)
local n = b10Cache[value]
if n then
return n
end
n = 0
for i = 1, #value do
n = n + 93 ^ (i - 1) * dictionary[sub(value, -i, -i)]
end
b10Cache[value] = n
return n
end
local function compress(text)
local dictionaryCopy = copy(dictionary)
local key, sequence, size = "", {}, #dictionaryCopy
local width, spans, span = 1, {}, 0
local function listkey(k)
local value = tobase93(dictionaryCopy[k])
local valueLength = #value
if valueLength > width then
width, span, spans[width] = valueLength, 0, span
end
insert(sequence, rep(" ", width - valueLength) .. value)
span += 1
end
text = escape(text)
for i = 1, #text do
local c = sub(text, i, i)
local new = key .. c
if dictionaryCopy[new] then
key = new
else
listkey(key)
key = c
size += 1
dictionaryCopy[new], dictionaryCopy[size] = size, new
end
end
listkey(key)
spans[width] = span
return table.concat(spans, ",") .. "|" .. table.concat(sequence)
end
local function decompress(text)
local dictionaryCopy = copy(dictionary)
local sequence, spans, content = {}, string.match(text, "(.-)|(.*)")
local groups, start = {}, 1
for span in string.gmatch(spans, "%d+") do
local width = #groups + 1
groups[width] = sub(content, start, start + span * width - 1)
start = start + span * width
end
local previous
for width, group in ipairs(groups) do
for value in string.gmatch(group, rep(".", width)) do
local entry = dictionaryCopy[tobase10(value)]
if previous then
if entry then
insert(dictionaryCopy, previous .. sub(entry, 1, 1))
else
entry = previous .. sub(previous, 1, 1)
insert(dictionaryCopy, entry)
end
insert(sequence, entry)
else
sequence[1] = entry
end
previous = entry
end
end
return unescape(table.concat(sequence))
end
return { Compress = compress, Decompress = decompress }
I am specifically trying to get decompression to be as fast as possible
1 Like
Would Parallel Lua be at all a possibility here? I’m not sure. It’s the only thing I can think of other than just reducing the amount of data you’re trying to compress.
4 Likes
It could be, but I have no idea how I would even approach implementing that to this. Would buffers be useful at all?
1 Like
It think that buffers should make it faster, as everytime a function that manipulates string is used, it needs to create that string in memory. I may not be exactly right but that’s how i remember it worked. So it is possible that buffers would make it faster.
1 Like
It might not be possible if you compress the string as a whole, however, if the string can be split into segments, then yes, it is definitely possible. In one of my game, I have a big string that I decompress using my parallel lua module, but that big string has a separator (I used “_”), separating the data of the different games stored
I use string.split() to split the string, and then do some math to spread the work amongst a set number of actors. Removing some of the string manipulation steps could improve performance further, but this alone made it a lot quicker
This also allows me to update only a section of the giant string, meaning I don’t have to compress and decompress everything every time I update part of it
Another approach would be to have each “section” be a fixed length, which I don’t know how would work, but is definitely possible
One this that annoys me with this module is that it uses base 93, while I use base 92 for the serialization (it uses a modified version of squash). This is for the separator I use, mentioned above
2 Likes
So… after spending several days on this, trying to figure out if there’s a way to further optimize this.
I ended up bringing the compression speed 10-50% faster, on average, with buffers!
--!native
-- Module by 1waffle1 and boatbomber, optimized and fixed by iiau
-- https://devforum.roblox.com/t/text-compression/163637/37
local lzw = {}
lzw.__index = lzw
lzw.new = function(BUF_SIZE: number?)
local self = setmetatable({}, lzw)
self.BUF_SIZE = BUF_SIZE or 8388608 --* 8 MB should be enough for most things on Roblox
self.buf = buffer.create(self.BUF_SIZE)
self.strToLength = {}
self.lengthToStr = buffer.create(94) -- base93
self.bidirectionalDict = {}
local lengthToStr = self.lengthToStr
local strToLength = self.strToLength
local bidirectionalDict = self.bidirectionalDict
do -- populate dictionary
local length = 0
for i = 32, 127 do
if i ~= 34 and i ~= 92 then
local c = string.char(i)
strToLength[c] = length
bidirectionalDict[length] = c
bidirectionalDict[c] = length
buffer.writeu8(lengthToStr, length, i)
length = length + 1
end
end
end
local escapemap_126, escapemap_127 = {}, {}
local unescapemap_126, unescapemap_127 = {}, {}
local blacklisted_126 = { 34, 92 }
for i = 126, 180 do
table.insert(blacklisted_126, i)
end
do -- Populate escape map
-- represents the numbers 0-31, 34, 92, 126, 127 (36 characters)
-- and 128-180 (52 characters)
-- https://devforum.roblox.com/t/text-compression/163637/5
for i = 0, 31 + #blacklisted_126 do
local b = blacklisted_126[i - 31]
local s = i + 32
-- Note: 126 and 127 are magic numbers
local c = string.char(b or i)
local e = string.char(s + (s >= 34 and 1 or 0) + (s >= 91 and 1 or 0))
escapemap_126[c] = e
unescapemap_126[e] = c
end
for i = 1, 255 - 180 do
local c = string.char(i + 180)
local s = i + 34
local e = string.char(s + (s >= 92 and 1 or 0))
escapemap_127[c] = e
unescapemap_127[e] = c
end
end
local function escape(s : string)
-- escape the control characters 0-31, double quote 34, backslash 92, Tilde 126, and DEL 127 (36 chars)
-- escape characters 128-180 (53 chars)
return string.gsub(string.gsub(s, '[%c"\\\126-\180]', function(c)
return "\126" .. escapemap_126[c]
end), '[\181-\255]', function(c)
return "\127" .. escapemap_127[c]
end)
end
local function unescape(s : string)
return string.gsub(string.gsub(s, "\127(.)", function(e)
return unescapemap_127[e]
end), "\126(.)", function(e)
return unescapemap_126[e]
end)
end
local b93Cache = {}
local b10Cache = {}
local function tobase93(n: number) : string
local value : string? = b93Cache[n]
if value then
return value
end
local c = n
value = ""
repeat
local remainder = n % 93
value = string.char(buffer.readu8(lengthToStr, remainder)) .. value
n = (n - remainder) / 93
until n == 0
b93Cache[c] = value
return value
end
local function tobase10(value: string) : number
local n = b10Cache[value]
if n then
return n
end
n = 0
for i = 1, #value do
n = n + math.pow(93, i - 1) * strToLength[string.sub(value, -i, -i)]
end
b10Cache[value] = n
return n
end
self.escape = escape
self.unescape = unescape
self.tobase93 = tobase93
self.tobase10 = tobase10
return self
end
function lzw:compress(text: string)
assert(type(text) == "string", "bad argument #1 to 'compress' (string expected, got " .. typeof(text) .. ")")
local buf = buffer.fromstring((self.escape :: (string) -> string)(text))
local tobase93 = self.tobase93
local dictionaryCopy : {[string] : number} = table.clone(self.strToLength)
local sequence: buffer, size : number = self.buf, 93
local width: number, spans: {[number]: number}, span : number = 1, {}, 0
local ptrA: number, ptrB: number = 0, 1
local key: string = ""
local cursor: number = 0
buffer.fill(sequence, 0, 0)
local function listkey(k : string) -- string to length
local value = tobase93(dictionaryCopy[k])
local valueLength = #value
if valueLength > width then
width, span, spans[width] = valueLength, 0, span
end
for _ = 1, width - valueLength do
buffer.writeu8(sequence, cursor, 32)
cursor += 1
end
buffer.writestring(sequence, cursor, value)
cursor += valueLength
span += 1
end
local len = buffer.len(buf)
while ptrB <= len do
local new : string = buffer.readstring(buf, ptrA, ptrB - ptrA)
if dictionaryCopy[new] then
key = new
ptrB += 1
else
listkey(key)
ptrA = ptrB - 1
size += 1
dictionaryCopy[new] = size
end
end
listkey(key)
spans[width] = span
return table.concat(spans, ",") .. "|" .. buffer.readstring(sequence, 0, cursor)
end
function lzw:decompress(text: string)
assert(type(text) == "string", "bad argument #1 to 'decompress' (string expected, got " .. typeof(text) .. ")")
local dictionaryCopy = table.clone(self.bidirectionalDict)
local tobase10 = self.tobase10
local sequence, spans, content = {}, string.match(text, "(.-)|(.*)")
local groups, start = {}, 1
for span in string.gmatch(spans, "%d+") do
local width = #groups + 1
groups[width] = string.sub(content, start, start + span * width - 1)
start = start + span * width
end
local previous
for width, group in ipairs(groups) do
for value in string.gmatch(group, string.rep(".", width)) do
local entry = dictionaryCopy[tobase10(value)]
if previous then
if entry then
table.insert(dictionaryCopy, previous .. string.sub(entry, 1, 1))
else
entry = previous .. string.sub(previous, 1, 1)
table.insert(dictionaryCopy, entry)
end
table.insert(sequence, entry)
else
sequence[1] = entry
end
previous = entry
end
end
return self.unescape(table.concat(sequence))
end
return lzw
I’ve also turned this into moreso a class, where you create a new encoder and decoder with lzw.new(bufferSize? = 8mb).
Methods are:
lzw:compress(string)
lzw:decompress(string)
And that’s it!
Decompression remains unchanged - I believe that encoding will generally used more in practice (think: ProfileService autosaves).
How does this optimization save time?
Turns out, that table indexing and buffer accessing are the same. But, we can save time on buffer writing! This makes use of writing the sequence into a fixed block memory instead of through a vector, which doesn’t need to spend CPU work to internally resize the array. Also, it’s faster to index a buffer rather than index a substring of a string.
The default size is 8 MB which is more than enough for DataStores (maximum is 4mb).
This is backwards compatible with my last release in this thread.
2 Likes
I’ve optimized this even further! By directly pushing the characters into the buffer after computing the base-93 literals, I’m able to cut down the first time compression speed an additional 25-50% in most cases!
Small text (<1024) characters see the most improvement by nearly 60-80%!
The only drawback is that I’m unable to memoize the results of the base-93 calculation, as reading a string from the buffer for the sole purpose of storing it in a dictionary will mean significant slowdowns - meaning that it may be slower on repeat calculations than in the original implementation, especially on large inputs.
However, given that in most scenarios you have constantly evolving save data on players and different players - it won’t be a big issue!
Furthermore, optimizing the escape function with buffers also provided a small speed increase.
Edit: This has been fixed on 10/16/24! 10/17/24 
10/16/24
Line 171 just needed to be deleted from
cursor += sz
into
cursor += sz - 1
Also removed unnecessary buffer.fill.
Line 171 just needed to be deleted (cursor += sz + 1)
DO NOT USE THIS. See my reply below for a fixed version.
--!native
-- Module by 1waffle1 and boatbomber, optimized and fixed by iiau
-- https://devforum.roblox.com/t/text-compression/163637/37
local lzw = {}
lzw.__index = lzw
--local concatSize = 512
lzw.new = function(BUF_SIZE: number?)
local self = setmetatable({}, lzw)
self.BUF_SIZE = BUF_SIZE or 8388608 --* 8 MB should be enough for most things on Roblox
self.buf = buffer.create(self.BUF_SIZE)
--self.concat = buffer.create(concatSize) -- storage to concat strings (is this faster?)
--local concat = self.concat
self.strToLength = {}
self.lengthToStr = buffer.create(94) -- base93
self.bidirectionalDict = {}
local lengthToStr = self.lengthToStr
local strToLength = self.strToLength
local bidirectionalDict = self.bidirectionalDict
do -- populate dictionary
local length = 0
for i = 32, 127 do
if i ~= 34 and i ~= 92 then
local c = string.char(i)
strToLength[c] = length
bidirectionalDict[length] = c
bidirectionalDict[c] = length
buffer.writeu8(lengthToStr, length, i)
length = length + 1
end
end
end
local escapemap_126, escapemap_127 = {}, {}
local unescapemap_126, unescapemap_127 = {}, {}
local blacklisted_126 = { 34, 92 }
for i = 126, 180 do
table.insert(blacklisted_126, i)
end
do -- Populate escape map
-- represents the numbers 0-31, 34, 92, 126, 127 (36 characters)
-- and 128-180 (52 characters)
-- https://devforum.roblox.com/t/text-compression/163637/5
for i = 0, 31 + #blacklisted_126 do
local b = blacklisted_126[i - 31]
local s = i + 32
-- Note: 126 and 127 are magic numbers
local c = b or i
local e = s + (s >= 34 and 1 or 0) + (s >= 91 and 1 or 0)
escapemap_126[c] = e
unescapemap_126[string.char(e)] = string.char(c)
end
for i = 1, 255 - 180 do
local c = i + 180
local s = i + 34
local e = s + (s >= 92 and 1 or 0)
escapemap_127[c] = e
unescapemap_127[string.char(e)] = string.char(c)
end
end
local function escape(s : string)
local len = #s
local refbuf = buffer.fromstring(s)
local newbuf = buffer.create(len * 2) -- worst case scenario
local c = 1 -- 1 indexed
local f = string.find(s, '[%c"\\\126-\255]') -- 1 indexed
local bufcursor = 0
while f do
if c < f then
buffer.copy(newbuf, bufcursor, refbuf, c - 1, f - c)
bufcursor += f - c
c = f
end
local byte = buffer.readu8(refbuf, f - 1)
if byte >= 181 then
local e = escapemap_127[byte]
buffer.writeu8(newbuf, bufcursor, 127)
buffer.writeu8(newbuf, bufcursor + 1, e)
else
local e = escapemap_126[byte]
buffer.writeu8(newbuf, bufcursor, 126)
buffer.writeu8(newbuf, bufcursor + 1, e)
end
c += 1
bufcursor += 2
f = string.find(s, '[%c"\\\126-\255]', f + 1)
end
if c <= len then
buffer.copy(newbuf, bufcursor, refbuf, c - 1, len - c + 1)
bufcursor += len - c + 1
end
return newbuf, bufcursor
end
local function unescape(s : string)
return string.gsub(string.gsub(s, "\127(.)", function(e)
return unescapemap_127[e]
end), "\126(.)", function(e)
return unescapemap_126[e]
end)
end
local b10Cache = {}
local function tobase10(value: string) : number
local n = b10Cache[value]
if n then
return n
end
n = 0
for i = 1, #value do
n = n + math.pow(93, i - 1) * bidirectionalDict[string.sub(value, -i, -i)]
end
b10Cache[value] = n
return n
end
self.escape = escape
self.unescape = unescape
--self.tobase93 = tobase93
self.tobase10 = tobase10
return self
end
function lzw:compress(text: string)
assert(type(text) == "string", "bad argument #1 to 'compress' (string expected, got " .. typeof(text) .. ")")
local buf, len = self.escape(text)
local dictionaryCopy : {[string] : number} = table.clone(self.strToLength)
local sequence: buffer, size : number = self.buf, 93
local width: number, spans: {[number]: number}, span : number = 1, {}, 0
local ptrA: number, ptrB: number = 0, 1
local key: string = ""
local cursor : number = 0
local lengthToStr : buffer = self.lengthToStr
local depth : number
local writeasbase93 = function(n: number) : string
local sz = 0
repeat
local remainder = n % 93
buffer.copy(sequence, cursor + depth - 1 - sz, lengthToStr, remainder, 1)
n = (n - remainder) / 93
sz += 1
until n == 0
end
local function listkey(k : string) -- string to length
local n = dictionaryCopy[k]
depth = n <= 1 and 1 or math.ceil(math.log(n, 93))
if depth > width then
width, span, spans[width] = depth, 0, span
end
for _ = 1, width - depth do
buffer.writeu8(sequence, cursor, 32)
cursor += 1
end
writeasbase93(n)
cursor += depth
span += 1
end
while ptrB <= len do
local new : string = buffer.readstring(buf, ptrA, ptrB - ptrA)
if dictionaryCopy[new] then
key = new
ptrB += 1
else
listkey(key)
ptrA = ptrB - 1
size += 1
dictionaryCopy[new] = size
end
end
listkey(key)
spans[width] = span
return table.concat(spans, ",") .. "|" .. buffer.readstring(sequence, 0, cursor)
end
function lzw:decompress(text: string)
assert(type(text) == "string", "bad argument #1 to 'decompress' (string expected, got " .. typeof(text) .. ")")
local dictionaryCopy = table.clone(self.bidirectionalDict)
local tobase10 = self.tobase10
local sequence, spans, content = {}, string.match(text, "(.-)|(.*)")
local groups, start = {}, 1
for span in string.gmatch(spans, "%d+") do
local width = #groups + 1
groups[width] = string.sub(content, start, start + span * width - 1)
start = start + span * width
end
local previous
for width, group in ipairs(groups) do
for value in string.gmatch(group, string.rep(".", width)) do
local entry = dictionaryCopy[tobase10(value)]
if previous then
if entry then
table.insert(dictionaryCopy, previous .. string.sub(entry, 1, 1))
else
entry = previous .. string.sub(previous, 1, 1)
table.insert(dictionaryCopy, entry)
end
table.insert(sequence, entry)
else
sequence[1] = entry
end
previous = entry
end
end
return self.unescape(table.concat(sequence))
end
return lzw
3 Likes
One of the smartest things I’ve ever seen base-93 is a crazy hack.
Hello, I tried using this but kept having an error occur when the string length is above 1:
attempt to perform arithmetic (mul) on number and nil
This error is from the base10 function from when attempting to convert string to length.
n += math.pow(93, i - 1) * strToLength[string.sub(value, -i, -i)]
While I work on a fix, in the meantime just use the decode from this code.
EDIT Never mind. I see what’s going on. Turns out every character seems to have the \0 character in between every character when it’s encoded. I’ll see what’s going on.
EDIT 2 See my last post. It is fixed