iraf.net

Hello Marianne,> I have encountered this problem while running a script of mine:
>
> ---
> INTERNAL ERROR on line 73: dictionary full
> rspectext (input=fort.22, output=sp1d-dr-1285-52723-0558.fits)
> ----
>
> fort.22 is 3847 lines long - so nothing extreme.No matter what the scripting language, 3847 lines is actually quite
long. It is unfortunate that the CL has a static dictionary that can
overflow, but it isn't surprising that you're challenging the size of
the dictionary.The first thing to do is to consult the FAQ: http://iraf.noao.edu/
scripts/search.faq?pattern=dictionary&type=and, although I wouldn't
typically suggest recompiling the CL.Second, I note that you modified the script yesterday. Did it work
before you made the latest change?> It crashes after having processed 24 or 49 spectra (I tried running
> it twice with a long list of input spectra to process, one at a time).Is it possible to simply run the spectra in two batches?You might see if the problem goes away under the new Enhanced CL:
http://iraf.noao.edu/iraf/web/new_stuff/ecl_beta.htmlI hope some of these suggestions help.Rob Seaman
NOAO/IRAF

Hi Rob,
Thanks for your message. I have run it before but interactively
within the "continuum" task. I tried taking out the section which
does the "rspectext" part but then the program just crashes a few spectra
later. Besides, if the size of the input ascii file is the problem, why
does the program not crash right away but only after processing of order
50 spectra? Also, in this day and age, a length of a spectrum of 4000
pixels isn't unusual.The problem of running the program in batches is that I'll have
to rebuild all the variable names etc in each sub-program. Do you mean
that I run the spectra in batches of ~40 spectra? Well, I already have
46000 spectra to process in batches of several hundred spectra at a
time, so....What type of records is it typically which are a problem? I don't
have a lot of open list pointers.Thanks,
Marianne************************************************************************
* Marianne Vestergaard Steward Observatory *
* mvestergaard@as.arizona.edu University of Arizona *
* Phone: 520-621-5950 933 N. Cherry Avenue *
* Telefax: 520-621-1532 Tucson, AZ 85721, USA *
* http://gloegg.as.arizona.edu/~vester/ *
************************************************************************
On Thu, 20 Oct 2005, Rob Seaman wrote:> Hello Marianne,
>
>> I have encountered this problem while running a script of mine:
>>
>> ---
>> INTERNAL ERROR on line 73: dictionary full
>> rspectext (input=fort.22, output=sp1d-dr-1285-52723-0558.fits)
>> ----
>>
>> fort.22 is 3847 lines long - so nothing extreme.
>
> No matter what the scripting language, 3847 lines is actually quite long. It
> is unfortunate that the CL has a static dictionary that can overflow, but it
> isn't surprising that you're challenging the size of the dictionary.
>
> The first thing to do is to consult the FAQ:
> http://iraf.noao.edu/scripts/search.faq?pattern=dictionary&type=and, although
> I wouldn't typically suggest recompiling the CL.
>
> Second, I note that you modified the script yesterday. Did it work before
> you made the latest change?
>
>> It crashes after having processed 24 or 49 spectra (I tried running
>> it twice with a long list of input spectra to process, one at a time).
>
> Is it possible to simply run the spectra in two batches?
>
> You might see if the problem goes away under the new Enhanced CL:
> http://iraf.noao.edu/iraf/web/new_stuff/ecl_beta.html
>
> I hope some of these suggestions help.
>
> Rob Seaman
> NOAO/IRAF

Hi Marianne,> Also, in this day and age, a length of a spectrum of 4000 pixels
> isn't unusual.Sorry - I think we're talking at cross purposes. I understood you to
say that your script file - the program, not the spectra, that is -
was close to 4000 lines long. As you say, a 4000 pixel spectrum is
nothing special. IRAF tasks should be able to handle spectra many,
many thousands (if not millions) of pixels long.The dictionary full message has nothing to do with the size of the
spectra - or even the number of the spectra, rather with the size or
complexity of the script. Typically this might involve calling some
set of tasks repeatedly within a loop and for whatever reason the
tasks keep getting added to the end of the dictionary rather than
replacing previous entries.> Do you mean that I run the spectra in batches of ~40 spectra?That was the suggestion.> Well, I already have 46000 spectra to process in batches of several
> hundred spectra at a time, so....Are you familiar with "@-files" for specifying a list of images (or
spectra)? If the current version of your script is found to work on
batches of a few dozen spectra, you could divide the 46000 spectra up
into separate @-files of a few dozen each. Then populate a new
script with commands like:mytask @file1
mytask @file2
mytask @file3
...and run the whole thing as one batch. You might find that
interspersing the magic word "flpr" helps:mytask @file1
flpr
mytask @file2
flpr
...The bottom line is that I don't know why your script is now
overflowing the dictionary. It sounds like it worked before your
most recent change to the logic of the script. (?) I presume,
however, that you care more about processing the spectra instead of
debugging the script. Did you happen to try the enhanced CL as
suggested in the previous message?Rob