Mikhail Korobov

Hey @moonlet, sorry for a delay! Could you please generate C++ files in a separate commit? It is hard to review the PR because github doesn't show what is changed...

@hzitoun could you please try installing from github? I've updated DAWG to support recent Pythons, but haven't made a release yet.

Hi, This is unexpected: ``` >>> t['d'] 'c' >>> t['e'] 'c' ``` It should raise KeyError (test passes for me under python 2.6, 2.7, 3.2 and 3.3 on Mac OS...

Hi Dan, I think there could be 2 issues. 1. By default DAWG constructor "unrolls" generator and sorts it because the library requires data to be sorted before the insertion....

Once I was trying to store a huge number of bigrams, each bigram was associated with a float number. I was not interested in high precision, so the first approach...

Anyway, I don't know why DAWG saving didn't raise an exception; this looks like a bug.

This is likely happens because of lack of validation. You alphabet is 0-255, but the data contain values outside this alphabet. Unfortunately, libdatrie (which is this wrapper using) doesn't really...

I've taken a quick look at your code, and it seems you're building a trie at the startup and then just using it. If so, you may give https://github.com/kmike/DAWG a...

Even better than (3) is to talk to libdatrie author and make its "unicode support" optional. Currently, alphabet is a range of unicode chars that datrie support. It doesn't use...

Here is one more for your pleasure: https://github.com/kmike/hat-trie :) I was looking for a perfect data structure at one point of my life.. I agree it should raise an exception....