[Info-vax] Parsers (was Re: VMS Features I Wish Linux Had)

[Stephen Hoffman]

On 2016-06-14 13:32:45 +0000, Bob Koehler said:

>  I feel the same way.  Early on I looked at LIB$TPARSE and decided not 
> to try to learn it.  CLD was a piece of cake.

The lib$table_parse documentation is comparatively weak, and the use of 
macros and not a DSL is not something likely to be emulated.   The tool 
itself is easy, though quite primitive in terms of what you have to 
code to use it.  Lots of glue code.  The need for glue code for many of 
the OpenVMS APIs has been a complaint of mine for a while.   There are 
ways around that necessity now, ways that hadn't been common in the 
1970s; when the majority of the OpenVMS API patterns were conceived.  
e.g.  OO.  But I digress.  There are also limitations around the 
token-parsing grammar, too.   http://labs.hoffmanlabs.com/node/1905

If you're creating a DCL-style command, then certainly use the SET 
COMMAND tool and CLD files.   If you're solving a problem that does not 
lend itself to DCL syntax, then lib$table_parse or some other parsing 
tool or library will be far preferable and far more effective than 
using CLD files.

Use of flex/bison, byacc, Ragel, or other such tools are much more 
common these days, though.  Ragel http://labs.hoffmanlabs.com/node/1629 
and byacc http://invisible-island.net/byacc/byacc.html are already 
available on OpenVMS, and byacc features licensing that would likely 
allow its incorporation into OpenVMS itself, if VSI were so inclined.


-- 
Pure Personal Opinion | HoffmanLabs LLC