return code r(9611) as ado file but not as do file

[Darren White]

Hi Stata listers

Looking for insight into an odd error. The file below runs correctly when executed as a do from a do file window but returns

r(9611)

OR

foreach ifs1 in `if1' {
program 0while already defined
r(110);
r(110);

interchangeably when run as an ado file. The include command calls another do file which uses the locals (strv indv etc)

In the code I migrated the globals to locals from user input to splice with the existing code of the included do file and also to try and weed out the problem if it was related to global and local issues.

The included do file is not buggy and has worked for several months properly so I don't suspect that.
My hunch is that it is something to do with the translation from global to local and that the ado vs do execution has some subtle difference in the referencing of these macros?

Any insight would be appreciated.

cheers

Darren


************************************************** *************************************************

program cubeit

clear
macro drop _all

display"Enter Dataset type Adult=1 Child=2: "_request(rdstyp)
display"Enter dataset year: "_request(rdsyr)
display"Enter weight variable: "_request(rwtvar)
display"Enter categorical=1 or continuous=2 variables: "_request(rcatcon)
display"Enter vari list: "_request(rvaril)
display"Enter strat list: "_request(rstratl)
display"Enter if1 list (default is dumif): "_request(rif1l)
display"Enter if2 list (default is dumif): "_request(rif2l)
display"Enter if3 list (default is dumif): "_request(rif3l)
display"Enter geographies: "_request(rgeogl)
*migrate globals to locals and align names to the compiler code
local strv $rstratl
local indv $rvaril
local if1 dumif $rif1l
local if2 dumif $rif2l
local if3 dumif $rif3l
local dsyr $rdsyr
local geov $rgeogl
local date=c(current_date)
local dstyp=$rdstyp
local catcon=$rcatcon
*+"_"+ c(current_time)

macro list

if `dstyp'==1 {
use "Q:\STATA\Surveys\master_data\urfs\SRHS`dsyr'_Mast er", clear
}
*
if `dstyp'==2 {
use "Q:\STATA\Surveys\master_data\urfs\CHS`dsyr'_Maste r", clear
}
*
svyset [pweight=$rwtvar]
gen dumif=90
gen state=300000
label drop _all

if `catcon'==1 {
include "Q:\STATA\Surveys\est_db\CUBE 2.0\datacube\code\static_code\cube2compiler_catvar .do"
}
*
if `catcon'==2 {
include "Q:\STATA\Surveys\est_db\CUBE 2.0\datacube\code\static_code\cube2compiler_convar .do"
}
*
if `dstyp'==1 {
save "Q:\STATA\Surveys\est_db\cubecompiler\SRHS`dsyr'`d ate'"
}
*
if `dstyp'==2 {
save "Q:\STATA\Surveys\est_db\cubecompiler\CHS`dsyr'`da te'"
}

end
exit

[Carole J. Wilson]

Use trace to identify where the program is failing:

set more off
set trace on
cubeit

If the output is too verbose, set the depth of the trace much lower (2, 3, or 5):

set tracedepth 3

[Darren White]

Thanks Carole

I may have discovered the answer in the archive (below) And the trace function does show that the error triggers right at:

foreach ifs1 in `if1' {
program 0while already defined
r(110);


Very strange error. I will post if the suggested solution fixes the problem

Darren

__________________________________________________ _____________________________

From	"Brian Karfunkel" <[email protected]>
To	[email protected]
Subject	Re: st: Error r(9611)
Date	Fri, 17 Oct 2008 10:55:16 -0700

---

In trying to find out precisely what was causing the problem, I made some vary simple test programs that sort of mimic the structure of my original program: there is -run_test.do-, which calls the other programs, -test_loop.do-, which just does a simple for-loop, and -run_program.ado-, which simply tries to run/do/include -test_loop.do- (see complete code below). Running -run_test.do- causes error r(9611) just like my original program, but it does not have the problem with modifying the loop variable in the loop (which is a problem for other reasons anyway, though, so thanks for catching it). What I have concluded from my tests is that the error is thrown when an ADO-file -INCLUDE-s a do-file with a loop (any kind, foreach, forvalues, or while). The error is thrown at the moment the loop in the DO-file starts. Note that including the loop itself in the ADO-file doesn't cause an error, nor does -run-ing or -do-ing the DO-file with the loop. Indeed, I have developed a work-around for my original problem by simply copying the code I was trying to -include- and pasting it in the ado-file itself; so the code in question works when literally in the ado-file but not when it is in a do-file that is -include-d in the ado-file. It seemed fishy how capricious the error was to me, so I tried it on another Windows machine and on a UNIX machine using Stata IC, both of which also threw the error. Thanks for everyone's help again. CODE:

[Darren White]

No banana, the same code that is refered to in the " include" causes the same problem when pasted into the "include" position in the code above. In this configuration the ado wont run at all and
the r(9611) appears after you type "cubeit". The code at include "Q:\STATA\Surveys\est_db\CUBE 2.0\datacube\code\static_code\cube2compiler_catvar .do" is fine and runs fine from a do window (see attached code)

Stumped....

Any insights appreciated

Darren

[Nick Cox]

[P] error says

9xxx. Various messages, all indicating an unexpected system failure. You should never see such a message. If one occurs, save your data, and exit Stata immediately. Please email [email protected] to report the problem.

FWIW, I can't see any gain whatsoever from trying to convert this to a program. In various places you refer to a compiler but there is no Stata compiler. Your program would remain interpreted code even if it worked. As it wires in very specific details it's entirely appropriate as a .do file.

[wbuchanan]

Perhaps try adding:

 
cap prog drop cubeit

On the line immediately prior to the one with program cubeit on it. It could be an issue of the program still residing in memory without being dropped and reloaded.

[Nick Cox]

I think Stata is getting confused with file references. There is nothing in the program that seems to explain any reference to a program 0while and although I am speculating too the use of include is to me more suspicious. But you don't need to solve the problem. The do-file works fine as a do-file and recasting it as a program is pointless any way.

[Darren White]

Thanks Nick and Mr(?) Buchanan.

I will give the drop cubeit command a go to se if it works. It makes sense that there is something in memory but I have completely re-booted stata and re run the program with the same outcome so only a little hopeful.

Nick, I use the term compiler in a loose sense ( the code builds large blocks of estimates in Mata iteratively so I called it a compiler, probably a bad choice) .
The r(9xxx) warning is concerning. I will follow your advice and save it as a .do. Using "include" in a program is warned against in the documentation too ( committing multiple crimes) .

Thanks for your help and advice.

Darren

[wbuchanan]

Darren White since all of the parameters are passed interactively there isn't any reason to make it an .ado. If, however, users were asked to supply parameters on invocation it would be a bit different. Without knowing anything about the .do files that are being executed there isn't a ton that anyone could do to potentially help you refactor it into an executable program. In general, it seems that this isn't the best candidate for a program (e.g., highly specific script that calls other highly specific scripts). I'll generally avoid creating an .ado file when the use case is limited to only a single task (e.g., even if the task is running other specific tasks) and in a case like this would typically create a master .do file that I could use to selectively execute specific things.

[Darren White]

Thanks Mr Buchanan. I will run it as an .do. And thanks for the programming tips.

Darren