This is not a production repository. I'll use it for testing, benchmarking and playing with Avro Phonetic on Android Platform.

Expect horrible sphaggeti codes with zero comments, I didn't write them for your viewing pleasure. Period.

Forking will do no good unless you want to contribute.

1. Perst performs way better than sqlite! As Database I/O is significantly better now, I can skip in memory database. Small heap size restriction on Android (prior Honeycomb) is no more a concern.

2. Average word lookup still takes 300+ 150+ ms with java.util.regex on Galaxy Y. Need to check with different regex persers based on this benchmark.

   1. Rex (the fastest library mentioned in the benchmark) is discontinued,sites shows 404. I downloaded the library using code search engines, but it doesn't work, fails to match any pattern.
   2. ICU4J is blazing fast on average lookup (30ms vs. 150ms), but it's UnicodeSet class is not exactly a regex parser. If I choose to continue with that, all existing patterns may need to be changed. :-S
   3. monq.jfa.Regexp, dk.brics.automaton.RegExp, jregex.Pattern are either slower or shows near equal performance compared to built-in regex under Android. com.karneim.util.collection.regex.Pattern site shows 404. Discarded them all.

3. Need to check with a different approach with Perst OODBMS. Instead of linked word lookup, I'll place words right inside the pattern array. Should be some hundreds times faster, but db size may be huge! Update: Test done. DB size increased in tolerable limit, but database I/O performance gain was huge (20ms vs. 200 ms). Planned to use this.

4. Important: Pattern for 'o' needs to be adjusted when it is the last item. (Hint: remove ([ওোঅ]|(অ্য)|(য়ো?))?)