Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: error message about Mata memory


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: error message about Mata memory
Date   Mon, 23 Oct 2006 23:34:38 +0100

There is no argument here, despite any appearance
to the contrary. 

Note that the situation is complicated by two
extra factors, as far as I can see: 

1. Some of StataCorp's Mata code is proprietary 
and/or pre-compiled and thus hidden, so in that case
a full trace would benefit even the advanced user
little or nothing, except the exquisite pleasure of
being able to say "This appears to be StataCorp's bug, 
not mine". Otherwise put, if StataCorp 
introduces fuller tracing for Mata, I doubt that
will let you see inside proprietary code. 

2. In any case new or inexperienced users who call 
for full tracebacks -- or whatever it is that SAS does -- 
will gain nothing by this unless they also learn some 
Stata or Mata programming. Others may be able to help diagnose
their problem a little easier, but my guess is: less
than they would expect. 

Nick 
n.j.cox@durham.ac.uk 

Kit Baum
 
> Nick said
> 
> Kit deleted the posting by Shihe Fan to which
> my posting was the reply. Shihe's posting was clearly
> about Stata. Despite the name of the thread, Shihe
> made a comment picking up something Bobby Gutierrez
> said (about -xtmixed-, a Stata command).
> 
> The story with Mata is indeed different and something
> like tracing I presume to be a matter Mata masters
> and mistresses will have on their amenda. In fact, it is
> surely a priority for MataCorp.
> 
> 
> 
> Indeed, I did not repeat the posting by Mr Fan, but Mr Fan's concern  
> was that the error reporting was 'not specific' -- that is, errors  
> thrown by Mata (which generated the original posting) were 
> not linked  
> to a line in the Mata source code. That is indeed true, and 
> the error  
> trace provided by Stata for Mata errors is 'not specific', as I  
> mentioned---unlike the specificity that would arise from a program  
> written in C or Fortran, for instance. "set trace" is not 
> very useful  
> when ado-files invoke Mata functions. In Robby's response, 
> xtmixed (a  
> Stata command implemented as an ado-file) calls Mata functions, but  
> if an error occurs within the Mata function, little information is  
> available.

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index