The numbers represent an index into another array. The numbers are positive. I could offset the numbers by a fixed amount if it would help somehow.

I have about 50,000 numbers today, but I don't want the code to have a 16 bit limit since I expect the dataset to grow.

I want to be able to store the data with Storable, and I think that it will not support a custom Inline::C representation. If there is another high-performance way to store the data that can take advantage of Inline::C, then that would be a candidate.

My platform is somewhat crufty. It is HP-UX 10.20 for the production web server and linux as the data generator. For generality and future-proofing, I prefer portable code, but this is not a hard requirement.

So far, my main ideas are:

• Translate the numbers to base64 to make them more compact in string form.
• Reorder the data so that more common numbers are smaller and are therefore shorter in string form.
• Create and optimize a state machine to make a real compressor tuned for this special set of characters.

I have been reluctant to use our database for this task because the lists of numbers can get quite long. I have a VARCHAR limit of 255 characters, and I don't want to get into BLOBs. I could use another database such as SQLite which uses strings as the datatype, but I have almost no experience with this type of solution. I do plan on implementing a solution (for the larger problem) in SQLite for comparison purposes.

The larger problem is that I am investigating speed versus memory tradeoffs in on-line queries involving a many-to-many relationship. I have an *extremely* fast solution using Table::ParentChild, but it cannot be stored using the Storable module. This is because Table::ParentChild uses an XS data structure, which represents each parent-child relationship in four 32 bit integers. I can sacrifice a little speed and use less memory. To this end I have coded another very fast solution using hashes and arrays, and most of the memory it uses are in these strings of numbers. I think that the strings are still wasteful, since they represent only about 3 bits of information per digit and one bit or so of information per delimiter character.

I hope to write more about the larger problem in the future, but for now I find that the core of my solution relies on these strings of numbers. The compression problem sounded familiar, so I hoped that someone would have some ideas.

I have run across this problem before, where I wanted compression, but I have a line-oriented file and I want to be able to decompress just a single line at a time, when this line is located anywhere in the file. It seems like an ordinary problem in compression, but after searching for a few hours I didn't find anything, so I asked first in chat and then here.

Update:
I have found the classification of the larger problem that I am working on. It is a graph problem. My graphs are sparse and large, so I am using an adjacency list. The sets that I am compressing are rows in the adjacency list. This is described in chapter 8 of Mastering Algorithms with Perl, where the Graph modules are described.

It should work perfectly the first time! - toma

---

---

• Are you posting in the right place? Check out Where do I post X? to know for sure.
• Posts may use any of the Perl Monks Approved HTML tags. Currently these include the following:

  <code> <a> <b> <big> <blockquote> <br /> <dd> <dl> <dt> <em> <font> <h1> <h2> <h3> <h4> <h5> <h6> <hr /> <i> <li> <nbsp> <ol> <p> <small> <strike> <strong> <sub> <sup> <table> <td> <th> <tr> <tt> <u> <ul>

• Snippets of code should be wrapped in <code> tags not <pre> tags. In fact, <pre> tags should generally be avoided. If they must be used, extreme care should be taken to ensure that their contents do not have long lines (<70 chars), in order to prevent horizontal scrolling (and possible janitor intervention).
• Want more info? How to link or How to display code and escape characters are good places to start.