Java-压缩大字串

Question

In my Java application I get from some calculations a really long string (from really long string I mean of around 600000 characters or so). But I need to send this string to a client to process it, and for this reason I need the compressed string to be of maximum 1000 characters.

I have tried using GZIPOutputStream and Inflater and Deflater classes, and in the best case I got an output string of 300000 characters, which is great compression, but in my case it's not enough.

I have also tried compressing the string n times, but the ouput was larger than the previous one, so only one-time compressing was successful.

So, what do you suggest me to try?

Thank you.

Answer 1

I agree with @Peter Lawrey that, strictly with those requirements, it might be impossible to deliver such a big message to the client.

Anyway, I still suggest three possible solutions, depending on how flexible your requirements are:

1. If all of your input strings have a restricted vocabulary (it does not allow a free, random combination of letters, symbols and numbers, but it is restricted to a certain set of business words, identifiers and values), and also a simple grammar , you can try to design your own compress algorithm . Example:

input symbol compressed symbol ------------ ----------------- client 1 bill 2 date 3 amount 4 value 5 price 6 tax 7

If the grammar is simple but the vocabulary is not that restricted, you could perform an initial custom compression to compress the document's structure as much as you can, and then a second GZIP compression to compress the data.

And don't forget that you'll have to bundle the client application with the corresponding uncompressor.

Anyway, it's not an easy task, I admit it.

2. Deliver the response to the client application in streaming . If the protocol is HTTP, you could use Chunked Transfer Coding .

3. If everything else fails, you'll have to page the results and serve them to the client by pages on demand: The client makes a query, the server executes it and delivers just the first page of results. Then, the client may chose to read the next page.