[Mumps2Py:] mumps2py and vista.rtn

[Patrick Gallot <pgallot@xxxxxxxxxxxx>]

Howdy,
After yesterday's commits, mumps2py is able to parse all but 36 of 
vista.rtn's 24 thousand  routines.
But it takes 30-40 minutes to parse them all (on my machine anyway).

so I wrote a utility to split vista.rtn by package (into roughly 130 
packages).  Most of the 36 routines that trigger a parsing error
are in the kernel package. If the parser knew how to handle the ^[ ] 
construct, the number of error-triggering
routines would be halved.

(I also found a number of trivial syntax errors in the PSORELD1, 
SDWLRP4, ZISHMSU, and ZISG3 routines,
including mismatched parentheses, a period that should have been a 
comman and $SELECT() terminators of the
form ":1,[...]", instead of ",1:[...]". those I fixed in my copy of 
vista.rtn)

Today,  I extracted  all the entryrefs from all the parsable routines, 
and extracted from that what packages depended
on what other packages.   The ideal would have been a set of 
dependencies that could be drawn as an acyclic graph.
VistA's package dependencies are not an acyclic  graph.

That's not too surprising considering that the VistA package is a purely 
conceptual device rather than a language-reinforced feature.
It's easy to accumulate, over time, features that cross conceptual 
boundaries and thereby undermine organizational principles.

I'm going to take a short break from coding mumps2py so that I can 
re-read/skim a few of my books on software development so that I can 
renew my acquaintance with software (and database and development) 
organization principles.

Cheers,
-Patrick



---