Re: recommendation for a WinNT/95 parser generator

Michael Tiomkin (mtiomkin@iil.intel.com) wrote:


: I had very bad experience with MKS's lex (1996, NT 3.51). The most
: simple solution was to use flex on Unix, and the native compiler on NT
: (MS C). Their other tools (make NT look almost like Unix!-) are much
: better.


Here is the text, which I had wrote for a similar thread in another
newsgroup.




I have not used yacc. But I use gnu-distributed bison version 1.24
and flex version 2.4.7


Both flex and bison were improvements on lex and yacc, which were
designed for c-programs working on per process FILE* structures. They
contain numerous static global variables and #defines which make life
horrible.


flex - has an option to generate an object oriented lexer class. flex
could even work on c++ streams, though i had encountered some problem
with it. But bison doesn't have any option to produce object oriented
class. bison however optionally guarantees your program to be
reentrant.


My personal experience is that if you program for a multithreaded
environment, where one of the threads is meant to invoke parser or
lexer, the existences of static global variables and per-process file
handles are potential nuisances.


I would rather have a parser/lexer who works on the file handles which
are opened by the win32 handles.


unicode-input is another requirement. So far I have not found, any
parser generator which produce parser/lexer who accepts their tokens
as unicode characters(16 bits), or multibyte chars.


-koshy
--
<(o)> Koshy George, <(o)> http://www.serv.net/~kgeorge <(o)>




--

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