| Re: How to call an IDL procedure from a DLM [message #47686] |
Fri, 24 February 2006 08:27 |
Rick Towler
Messages: 821
Registered: August 1998
|
Senior Member |
|
|
Maurizio Tomasi wrote:
> Karl Schultz wrote:
>> On Thu, 23 Feb 2006 08:26:54 -0500, Haje Korth wrote:
>>
<snip>
> The log window is of course implemented using pure IDL. The problem is
> that the DLM should write a message in the log window for each FITS file
> that is loaded successfully. And in order to do this it should call a
> "write_log_message" procedure, which is obviously written in IDL
> (because it uses IDL widgets).
Could you not still refactor your code... You wouldn't even have to
return the data to IDL while you are reading the files, just the status.
You could have a setup routine which you pass all of the filenames, a
read_one routine which reads a single file and returns the status, and a
data_return routine which returns all of the data to IDL. There would
be other ways too. I think you could do this with little impact on your
applications performance.
I believe you could also possibly use the external widget API
(widget_stub and associated functions) to create an invisible widget
that your DLM issues events to (IDL_WidgetIssueStubEvent()). Since the
event struct only has a single long field that you define, you also need
to write a routine that takes that value (say an address) and gets the
status of your DLM function. All of this is in the EDG under internal
API/Widgets.
-Rick
|
|
|
|
| Re: How to call an IDL procedure from a DLM [message #47697 is a reply to message #47686] |
Fri, 24 February 2006 00:27  |
Maurizio Tomasi
Messages: 7
Registered: December 2005
|
Junior Member |
|
|
Karl Schultz wrote:
> On Thu, 23 Feb 2006 08:26:54 -0500, Haje Korth wrote:
>
>> Maurizio,
>> What you are trying to do is an interesting concecpt, but I don't think DLMs
>> are intended to work this way. At least I have never run across an example
>> for such use. Isn't the whole purpose of DLMs, call_external, etc. to call
>> code that is specifically NOT implemeneted in IDL? Why would you bother
>> writing a DLM if the functionality is already present in IDL?
>
> I'm thinking that he is doing part of his DLM processing in C, and then
> gets to a point where it would be Really Nice to call an IDL routine to
> finish it up because the IDL routine is at-hand and might be a pain to
> translate into C or whatever.
>
> It should be possible to refactor the code though to do some things in C
> and then return the intermediate results to IDL, where he can then call
> the IDL code. I can't speculate further without knowing more about the
> problem.
My IDL program implements a full graphical user interface using the IDL
widget library, and is getting quite large (~ 80,000 lines of code). It
must load a large number (~ 5000) of FITS files whose size sums up to 1
GB. We cannot alter these numbers. The first version of the program was
completely written in IDL, and was really slow when loading files. Se we
are developing a DLM with C and the CFITSIO library, in order to make
the loading process faster (our preliminary tests show a gain in time of
more than one order of magnitude).
The program has a "log window" which keeps track of what the program is
doing. The purpose of this window is to have a record of what it has
been done. It is very important that every message is recorded in a file
as well, and that the file is immediately flushed and closed after each
message, in order for the messages to be saved even if the system breaks
down during the loading process.
The log window is of course implemented using pure IDL. The problem is
that the DLM should write a message in the log window for each FITS file
that is loaded successfully. And in order to do this it should call a
"write_log_message" procedure, which is obviously written in IDL
(because it uses IDL widgets).
Thank you for your suggestions,
Maurizio.
|
|
|
|
| Re: How to call an IDL procedure from a DLM [message #47705 is a reply to message #47697] |
Thu, 23 February 2006 09:17 |
Haje Korth
Messages: 651
Registered: May 1997
|
Senior Member |
|
|
Karl,
thanks for the info. So it seems even you folks at RSI think that this is
unfeasible. I usually blame things first on my C-illiteracy... :-)
"Karl Schultz" <k____schultz@rsinc.com> wrote in message
news:pan.2006.02.23.16.47.51.78000@rsinc.com...
> On Thu, 23 Feb 2006 08:26:54 -0500, Haje Korth wrote:
>
>> Maurizio,
>> What you are trying to do is an interesting concecpt, but I don't think
>> DLMs
>> are intended to work this way. At least I have never run across an
>> example
>> for such use. Isn't the whole purpose of DLMs, call_external, etc. to
>> call
>> code that is specifically NOT implemeneted in IDL? Why would you bother
>> writing a DLM if the functionality is already present in IDL?
>
> I'm thinking that he is doing part of his DLM processing in C, and then
> gets to a point where it would be Really Nice to call an IDL routine to
> finish it up because the IDL routine is at-hand and might be a pain to
> translate into C or whatever.
>
> It should be possible to refactor the code though to do some things in C
> and then return the intermediate results to IDL, where he can then call
> the IDL code. I can't speculate further without knowing more about the
> problem.
>
>> My guess is that you probably have to go out and find an external C or
>> FORTRAN library that does what you want, e.g. Numerical Recipes. RSI
>> manuals
>> often specify the source of their external algorithms. Or, if the
>> algortihm
>> is written in IDL, you can just translate it.
>>
>> This may just be well above my head, so maybe Karl has some input on
>> this. If this should really work, I would be interested in seeing an
>> example. I would be afraid that each of these calls need to start an
>> independent IDL session, thus leading to significant slowdown.
>
> You are on the right track. It isn't so much starting a new session, but
> rather reentering the interpreter that is causing the problem.
>
> I think redesigning the DLM as hinted above is the best approach. One way
> to "call IDL" from a DLM would be to spawn a new IDL process. But that
> would involve passing data back and forth via files or pipes/sockets and I
> think that's going to be more involved than refactoring the DLM.
>
> Karl
>
>
>>
>> Cheers,
>> Haje
>>
>> "Maurizio Tomasi" <tomasi@lambrate.inaf.it> wrote in message
>> news:dthqki$eh1$1@muffin.area.ba.cnr.it...
>>> Last week I posted a request for help, since my DLM produced a number
>>> of "Stopped on unknown instruction" errors. After some debugging, I
>>> found that the problem resides in IDL_ExecuteStr, which is not usable
>>> in DLMs since it has been meant to be used only with Callable IDL.
>>>
>>> However, I still need to call an IDL procedure from my DLM. I looked
>>> for a IDL_CallProc function, but I have had no success. Does somebody
>>> know if there is a way to do this?
>>>
>>> Thank you very much,
>>> Maurizio Tomasi.
>
|
|
|
|
| Re: How to call an IDL procedure from a DLM [message #47706 is a reply to message #47705] |
Thu, 23 February 2006 09:14 |
Haje Korth
Messages: 651
Registered: May 1997
|
Senior Member |
|
|
Rick,
I just looked at this as a prove of concept, not as something I would try
really hard to make use of. Life is hard enough as it is! :-)
Haje
"Rick Towler" <rick.towler@nomail.noaa.gov> wrote in message
news:dtkps4$jop$1@news.nems.noaa.gov...
>
> Maurizio Tomasi wrote:
>> Last week I posted a request for help, since my DLM produced a number of
>> "Stopped on unknown instruction" errors. After some debugging, I found
>> that the problem resides in IDL_ExecuteStr, which is not usable in DLMs
>> since it has been meant to be used only with Callable IDL.
>>
>> However, I still need to call an IDL procedure from my DLM. I looked for
>> a IDL_CallProc function, but I have had no success. Does somebody know if
>> there is a way to do this?
>
> I don't know if there is a way to do this but why do it the hard way? Why
> don't you wrap your dlm's functions in an IDL function or object and call
> the IDL procedures from within IDL?
>
> -Rick?
|
|
|
|
| Re: How to call an IDL procedure from a DLM [message #47708 is a reply to message #47706] |
Thu, 23 February 2006 08:47 |
Karl Schultz
Messages: 341
Registered: October 1999
|
Senior Member |
|
|
On Thu, 23 Feb 2006 08:26:54 -0500, Haje Korth wrote:
> Maurizio,
> What you are trying to do is an interesting concecpt, but I don't think DLMs
> are intended to work this way. At least I have never run across an example
> for such use. Isn't the whole purpose of DLMs, call_external, etc. to call
> code that is specifically NOT implemeneted in IDL? Why would you bother
> writing a DLM if the functionality is already present in IDL?
I'm thinking that he is doing part of his DLM processing in C, and then
gets to a point where it would be Really Nice to call an IDL routine to
finish it up because the IDL routine is at-hand and might be a pain to
translate into C or whatever.
It should be possible to refactor the code though to do some things in C
and then return the intermediate results to IDL, where he can then call
the IDL code. I can't speculate further without knowing more about the
problem.
> My guess is that you probably have to go out and find an external C or
> FORTRAN library that does what you want, e.g. Numerical Recipes. RSI manuals
> often specify the source of their external algorithms. Or, if the algortihm
> is written in IDL, you can just translate it.
>
> This may just be well above my head, so maybe Karl has some input on
> this. If this should really work, I would be interested in seeing an
> example. I would be afraid that each of these calls need to start an
> independent IDL session, thus leading to significant slowdown.
You are on the right track. It isn't so much starting a new session, but
rather reentering the interpreter that is causing the problem.
I think redesigning the DLM as hinted above is the best approach. One way
to "call IDL" from a DLM would be to spawn a new IDL process. But that
would involve passing data back and forth via files or pipes/sockets and I
think that's going to be more involved than refactoring the DLM.
Karl
>
> Cheers,
> Haje
>
> "Maurizio Tomasi" <tomasi@lambrate.inaf.it> wrote in message
> news:dthqki$eh1$1@muffin.area.ba.cnr.it...
>> Last week I posted a request for help, since my DLM produced a number
>> of "Stopped on unknown instruction" errors. After some debugging, I
>> found that the problem resides in IDL_ExecuteStr, which is not usable
>> in DLMs since it has been meant to be used only with Callable IDL.
>>
>> However, I still need to call an IDL procedure from my DLM. I looked
>> for a IDL_CallProc function, but I have had no success. Does somebody
>> know if there is a way to do this?
>>
>> Thank you very much,
>> Maurizio Tomasi.
|
|
|
|
| Re: How to call an IDL procedure from a DLM [message #47710 is a reply to message #47708] |
Thu, 23 February 2006 08:24 |
Rick Towler
Messages: 821
Registered: August 1998
|
Senior Member |
|
|
Maurizio Tomasi wrote:
> Last week I posted a request for help, since my DLM produced a number of
> "Stopped on unknown instruction" errors. After some debugging, I found
> that the problem resides in IDL_ExecuteStr, which is not usable in DLMs
> since it has been meant to be used only with Callable IDL.
>
> However, I still need to call an IDL procedure from my DLM. I looked for
> a IDL_CallProc function, but I have had no success. Does somebody know
> if there is a way to do this?
I don't know if there is a way to do this but why do it the hard way?
Why don't you wrap your dlm's functions in an IDL function or object and
call the IDL procedures from within IDL?
-Rick?
|
|
|
|
| Re: How to call an IDL procedure from a DLM [message #47716 is a reply to message #47710] |
Thu, 23 February 2006 05:26 |
Haje Korth
Messages: 651
Registered: May 1997
|
Senior Member |
|
|
Maurizio,
What you are trying to do is an interesting concecpt, but I don't think DLMs
are intended to work this way. At least I have never run across an example
for such use. Isn't the whole purpose of DLMs, call_external, etc. to call
code that is specifically NOT implemeneted in IDL? Why would you bother
writing a DLM if the functionality is already present in IDL?
My guess is that you probably have to go out and find an external C or
FORTRAN library that does what you want, e.g. Numerical Recipes. RSI manuals
often specify the source of their external algorithms. Or, if the algortihm
is written in IDL, you can just translate it.
This may just be well above my head, so maybe Karl has some input on this.
If this should really work, I would be interested in seeing an example. I
would be afraid that each of these calls need to start an independent IDL
session, thus leading to significant slowdown.
Cheers,
Haje
"Maurizio Tomasi" <tomasi@lambrate.inaf.it> wrote in message
news:dthqki$eh1$1@muffin.area.ba.cnr.it...
> Last week I posted a request for help, since my DLM produced a number of
> "Stopped on unknown instruction" errors. After some debugging, I found
> that the problem resides in IDL_ExecuteStr, which is not usable in DLMs
> since it has been meant to be used only with Callable IDL.
>
> However, I still need to call an IDL procedure from my DLM. I looked for a
> IDL_CallProc function, but I have had no success. Does somebody know if
> there is a way to do this?
>
> Thank you very much,
> Maurizio Tomasi.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
