Re: flex scanner too huge, suggestions?

I strongly suggest you try using re2c instead of flex. I just converted a
flex scanner over to re2c, and the size difference is stunning.


Re2c can also easily generate reentrant C lexers which is great. Re2c
requires you to write your own buffer management, but I copied the example
they provided for a C scanner and it was terrific.


re2c -b : c file size: 25470 (!!)
flex -f (%option align): c file size: 421804


Executable with re2c: 106252 bytes
Executable with flex: 325850 bytes


Re2c has some limitations on regular expression support, but the most
important features are there. The examples provided are very helpful.


Typically re2c generates faster and smaller parsers than flex. In my case,
the speed was about on par with flex. However, I got line and column
information support in the re2c scanner.


I wish I had found re2c sooner.


-Clint


re2c
----


Version 0.9.1
Originally written by Peter Bumbulis (peter@csg.uwaterloo.ca)
Currently maintained by Brian Young (bayoung@acm.org)


The re2c distribution can be found at:


        http://www.tildeslash.org/re2c/index.html


The source distribution is available from:


        http://www.tildeslash.org/re2c/re2c-0.9.1.tar.gz


>Made with LFLAGS = -Cfe, size gives
>42068(.text) + 24(.data) + 60(.bss) + 1687886(.rodata) = 1730038
>
>Made with LFLAGS = -Cem, size gives
>42496(.text) + 24(.data) + 60(.bss) + 508326(.rodata) = 550906
>
>This is still pretty large for some of our target environments.

Return to the comp.compilers page.
Search the comp.compilers archives again.