Re: st: Plugins, another strange behavior

[Kevin Crow <kcrow@stata.com>]

Sergiy reported a possible bug with Stata plugins in regards to missing values. 
 I could not reproduce the bug using a C plugin.

******************LOG**********************
. cap program drop printout

. program printout, plugin

.
. use sergiy, clear

. list

     +-----------------+
     | w   r   x     y |
     |-----------------|
  1. | 1   1   1   101 |
  2. | 1   1   2   102 |
  3. | 1   1   3   103 |
  4. | 1   1   0     . |
  5. | 1   1   5   105 |
     |-----------------|
  6. | 1   1   6   106 |
  7. | 1   1   7   107 |
  8. | 1   1   8   108 |
  9. | 1   1   9   109 |
 10. | 1   1   .   100 |
     +-----------------+

. plugin call printout w r x y
The value of x[1] is 1.000000
The value of x[2] is 2.000000
The value of x[3] is 3.000000
The value of x[4] is 0.000000
The value of x[5] is 5.000000
The value of x[6] is 6.000000
The value of x[7] is 7.000000
The value of x[8] is 8.000000
The value of x[9] is 9.000000
The value of x[10] is 898846567431157953864652595394512366808988489471153286367
*******************************************

******************C-Code**********************
#include <stdio.h>
#include "stplugin.h"

STDLL stata_call(int argc, char *argv[])
{
    char buf[80] ;
    ST_int  j ;
    ST_double z ;

        for(j = SF_in1(); j <= SF_in2(); j++) {
                if(SF_ifobs(j)) {
            (void) SF_vdata(3, j, &z) ;
            snprintf(buf, 80, "The value of x[%d] is %lf\n", j, z) ;
            SF_display(buf) ;
                }
        }
    return(0) ;
}
**********************************************

Sergiy, it might be best if you emailed me at tech-support@stata.com so that we 
can look into the problem further.

Kevin Crow - StataCorp

Sergiy Radyakin wrote:

> Hello Stata users,
>
> the following strange issue was discovered while writing a plugin to
> process the dataset currently in Stata's memory. It seems that
> casewise deletion of observations with missing values of variables
> passed on to the plugin is forced by Stata.
>
> Consider the following example:
>
>        w   r   x     y
>   1.   1   1   1   101
>   2.   1   1   2   102
>   3.   1   1   3   103
>   4.   1   1   0     .
>   5.   1   1   5   105
>   6.   1   1   6   106
>   7.   1   1   7   107
>   8.   1   1   8   108
>   9.   1   1   9   109
>  10.   1   1   .   100
>
>
> Variables x and y each have a missing value, though in different observations.
>
> An extract from the code (Delphi) which illustrates the problem follows:
>
> for i:=SF_in1 to SF_in2 do if SF_ifobs(i)=bTrue then
>            SF_Display(PChar('Processing observation: '+IntToStr(i)+'{break}'));
>
> The code loops through all observations from lower "in" value to upper
> "in" value and if the observation satisfies the "if" condition it's
> index is sent to Stata's output window.
>
> As a result we have:
>
> Processing observation: 1
> Processing observation: 2
> Processing observation: 3
> Processing observation: 5
> Processing observation: 6
> Processing observation: 7
> Processing observation: 8
> Processing observation: 9
>
> Where observations 4 (the one with the missing y) and 10 (the one with
> the missing x) are not listed. No "if"-condition was specified, and
> according to the manual for plugin-developers the function SF_ifobs()
> must return "True" for all observations.
>
> Both Stata 9 and Stata 10 show identical behavior in this issue. Is
> this supposed to be so? Is it controllable? How can one access x[4]
> and y[10] in this example?
>
> Thank you,
>     Sergiy Radyakin
>     Consultant, DECRG, The World Bank.
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/