[plt-scheme] WXME Decoding Error

From: Eric Domeshek (domeshek at stottlerhenke.com)
Date: Tue Oct 14 12:08:42 EDT 2008


Thanks for the suggestion on my WMXE file-reading problem.  In the end, 
I seem to have squashed (most of) that bug by more careful study of 
additional parts of the Framework docs.  I hadn't read the Handler 
chapter, so I was getting a default editor.  Now I've got appropriate 
calls to handler:insert-format-handler and 
handler:current-create-new-window so I get the right kind of window, 
which gives me my pasteboard% subclass as expected (and thus eliminates 
the unexpected text% editor). 

I've still got a few residual questions:

1- I also found the Finder chapter, and have calls to set up 
finder:default-extension and finder:default-filters.  However, these 
seem to only affect the Save/Save-As dialogs.  The Open dialog still 
comes up with "Any" as its only filter.  Is there a comparably simple 
way to set default filters on that dialog as well?

2- After my now (mostly) successful file load, I'm still getting the 
following error message:
> send: no such method: set-position for class: eer-pasteboard%
Based on searching the docs, this method seems to be associated with 
scroll-event%s and text% editors.  So maybe there's somewhere else in 
the framework that is still expecting a text% object?  Any thoughts on 
what else I might have overlooked as a PLT GUI newbie?

3- Let me take another shot at the other query I threw into my initial 
email: Where should I be stashing, and how should I be accessing, global 
(for a given editor's diagram or graph structure) state-objects, such 
that the specialized "read" methods of my specialized snip-class%s can 
find that state-object (say, in order to log the newly created snip%s 
into that object, or look up previously created snip%s or other 
underlying application objects that a new snip is supposed to connect up 

My current (bogus-feeling) approach is to count on the read method being 
called after on-file-load has run (and before any other on-file-load has 
run) so I can set up a global variable with a state-object for the 
file's contents.  As I said, this feels bogus, but I don't see how to 
make such shared data accessible to the read methods of these 
snip-class%s...  I started down the header/footer path at one point, but 
there's really no more data that needs to be stored in the file 
format--what I need to know is all part of (of reconstructible from) the 
snip data.

I suppose after the read methods complete there is an implicit call to 
the editor%'s insert method.  Is that the place to put any such 
editor-state-dependent code?  That seems a little funky, as I think of 
the state-management as being part of the creation of a 
complete/consistent snip% object (which seems like the job of read).  
I'd also have to overhaul the editor-actions that create the snip%s so 
that they rely on the same magic insert behavior to complete the snip% 
data structures.  Then of course there's figuring out if what I really 
want is to tack this onto on-insert or after-insert, and what to do 
about edit-sequences...  Somehow I was hoping for a simpler solution.

4- One final info-access style question: What is the right way to figure 
out the directory from which a module is being loaded?  My problem here 
is that I want my custom editor app to use some custom icons, so I've 
got an icons sub-directory beneath the directory that holds my code, but 
I don't know how to establish dynamically the directory that the code is 
loading from.  This seems like there must be a simple straightforward call.

Oh yes... I doubt it matters for these queries, but I forgot to say in 
my original note: I'm working with PLT 4.1.1 on Windows XP.

Thanks as always for the great support here.


Matthew Flatt wrote:
> At Fri, 10 Oct 2008 17:42:36 -0400, Eric Domeshek wrote:
>> But today I seem to have run into a bit of a brick wall in attempting to 
>> use the WXME encoding/decoding piece of the framework to save and reload 
>> my diagrams.  I've gotten a far as saving the data (writing a file with 
>> seven types of custom snips).  Unfortunately, I can't load the data back 
>> in, and the error message/trace is not helping me much.  Here's what 
>> comes up in the console window when I try to load one of these saved 
>> files by invoking the default File|Open command:
>>> insert-file in text%: error loading the file
>>> [...]
>> Can anyone suggest a way forward?  For starters, I'm not sure why 
>> there's a text% floating around.  I've only been dealing with 
>> pasteboard%s. 
> I don't have any great ideas, but you might try parsing the file using
> the `wxme' library, which might give better information about decoding
> problems.
> I'm puzzled by the "insert-file in text%" message, too. Looking at the
> source where that message can be reported, it seems to be always within
> a `text%' object, and never in a `pasteboard%'.
> If you're able and willing to send me code so I can see the problem on
> my machine, I would try it out.
>> Also, while I'm asking questions... As I read in my data, I need to 
>> create some custom structures that get wrapped inside my snips (i.e. 
>> "nodes" and "links" in an underlying graph structure).  That custom data 
>> all needs to be attached to a diagram-level object (the "graph").  But I 
>> don't see how the read methods in my various snip-class%es can best get 
>> hold of the diagram object (which is stashed in the frame%, which holds 
>> the pasteboard%, which will contain the snip%s that get created by those 
>> snip-class% read methods).  For now, I've got a pasteboard% on-load-file 
>> method setting a global variable so my various read methods can get hold 
>> of that shared diagram.  This seem ugly.  Any better ideas?
> I'm not sure I understand, but I wonder whether the information should
> instead reside in the pasteboard so that it's written and restored
> though `write-editor-global-footer' and `read-editor-global-footer'?
> Matthew

Posted on the users mailing list.