Comments on: LZW Data Compression

By: Mark Nelson

Mark Nelson — Tue, 01 May 2012 17:59:15 +0000

Hi Robert,

Sorry, I didn't get a chance to address this properly until today - I was traveling.

First, I think using pseudocode was a bad idea on my part, but it was a long time ago, and that was a different me.
I've got an updated version of the article that I think is a lot more trustworthy, and it only uses C++ code - no pseudocode. The decompress routine looks like this:

PLAIN TEXT

unsigned int code;
unsigned int next_code = 257;
while ( in>> code ) {
out <[code];
if ( previous_string.size() )
strings[next_code++] = previous_string + strings[code][0];
previous_string = strings[code];
}

In this case, previous_string is a standin for OLD_CODE, and it does exactly what you are saying it should do.

In the pseudo code, I am very casually going back and forth between strings and codes, which is a confusing mistake. So when I say OLD_CODE = NEW_CODE, I am doing so on the idea that NEW_CODE == STRING. Which it is, more or less, if you look up NEW_CODE in the library, you get STRING.

In any case, your correction is well taken, and as I say, I think the whole thing is a lot more clear and less prone to misinterpretation in the new article at:

http://marknelson.us/2011/11/08/lzw-revisited/

Thanks for your input, and thanks for reading.

- Mark

By: Robert K

Robert K — Mon, 30 Apr 2012 02:19:25 +0000

How does that work with line 8 then? "add OLD_CODE + CHARACTER to the translation table"

OLD_CODE needs to be a string in this context, no? Sure seems like what you want is the old string value in this case, not the old (unresolved) lookup code.

By: Mark Nelson

Mark Nelson — Sun, 29 Apr 2012 23:57:10 +0000

No.

Codes are integer values, which mostly start at 256 or 257 and increase montonically.

Strings are sequences of bytes.

So the assignment you propose doesn't work - it is assuming some kind of implicit conversion from one type to another, and that doesn't happen.

- Mark

By: Robert K

Robert K — Sun, 29 Apr 2012 22:43:03 +0000

I believe Line 9 of Figure 3 (decompression pseudocode):

PLAIN TEXT

OLD_CODE = NEW_CODE

... should actually be:

PLAIN TEXT

OLD_CODE = STRING

No?

By: lzw算法

lzw算法 — Wed, 11 Apr 2012 06:58:31 +0000

[...] COMPRESSION》作者Mark Nelson的文章lzw data compression中介绍为：LZW compression replaces strings of characters with single codes. It does not do any [...]

By: Mark Nelson

Mark Nelson — Wed, 07 Mar 2012 13:41:40 +0000

@Usha:

If you need to send the table along with the indices, you aren't doing LZW, you are doing something else.

- Mark

By: Usha

Usha — Wed, 07 Mar 2012 13:40:10 +0000

Hi Mark sir,
I have implemented LZW algorithm in MATLAB. I have transformed the message file (say 1MB) to a coding table of code words(1.4MB) and a table of indexes(50KB) which points to position of the code word in decompressed file. I want to send the compressed data i.e Indexes file on a wireless device. At the receiving end, how are we able to get the data stored in the coding Table? If we send that table also along with Indexes, size becomes double that of message file. Please clarify.

Thanks in advance

By: www.Phototalks.idv.tw » LZW compression & decompression

www.Phototalks.idv.tw » LZW compression & decompression — Mon, 02 Jan 2012 07:48:13 +0000

[...] 期末考記錄用參考自此但我覺得需要修改 [...]

By: Mark Nelson

Mark Nelson — Mon, 03 Oct 2011 20:03:51 +0000

@Patrick:

Isn't that a question for the people who make Lan Messenger? I have no idea.

By: Patrick

Patrick — Mon, 03 Oct 2011 19:57:51 +0000

Hi Mark, I have one question for you, I hope you could help me with this.

It is possible to embed LZW algorithm in LAN Messenger because I want to Compress and Decompress file and embed it in LAN Messenger? I want to use a Visual Basic or Visual C# programming language.

I hope you would reply immediately. Thank you.