Re: Change of current context ? (whos -file ...)

[John W. Eaton]

On 22-Sep-2008, David Bateman wrote:

| John W. Eaton wrote:
| 
| > I understand the motivation for this approach as it would be a
| > relatively simple change, but maybe we should instead modify do_load
| > in load-save.cc to skip the call to install_loaded_variable and return
| > the variable names instead?  I see do_load already has a way to return
| > the names (the list_only option).  This change might require a little
| > refactoring, but maybe it would not be too difficult?
| > 
| 
| In fact this is the way I tried to do it when I added the -regexp option 
| to the who function a few months ago. The logical way to treat this 
| would be to move the variables.cc (class symbol_info_list) to 
| variables.h and then use it i load-save.cc (do_load). This allows 
| exactly the same format as with whos itself in the load function.
| 
| However, the symbol_info_list class is in fact a list of symbol_records 
| and so there is in fact no gain in not calling install_loaded_variable 
| to a temporary context. I suppose the proper way is to rewrite 
| symbol_info_list to not store the symbol_record itself but only the 
| necessary information, though that is a bit more that a little bit of 
| refactoring.
| 
| The method I talked about doesn't let "load -l" profit from the same 
| code, but that is an Octave extension in any case, and so probably 
| better to use the Matlab way in any case. So in short for the minimum 
| effort its probably better just to do it the way I suggested but include 
| a FIXME comment discussing the issue.

OK, can you send me your preliminary changes as a diff that I could
apply to look at the scoping problem?  What is the symptom of the
problem?  How does it fail?

jwe