command table with option markvar(newvar) creates error __000001 not found

[Anders Alexandersson]

Why do the commands table and table twoway with the option markvar(newvar)create the error "__000001 not found r(111);"? The error occurs in Stata/MP 18.0 for Windows (64-bit x86-64) and also in Stata/MP 17.0 on the same platform.

. use https://www.stata-press.com/data/r18/nhanes2l, clear
(Second National Health and Nutrition Examination Survey)

. table sex, markvar(mynewvar)

---------------------
         |  Frequency
---------+-----------
Sex      |           
  Male   |      4,915
  Female |      5,436
  Total  |     10,351
---------------------
__000001 not found
r(111);

. noisily table sex diabetes, markvar(mynewvar2)

--------------------------------------------
         |           Diabetes status        
         |  Not diabetic   Diabetic    Total
---------+----------------------------------
Sex      |                                  
  Male   |         4,698        217    4,915
  Female |         5,152        282    5,434
  Total  |         9,850        499   10,349
--------------------------------------------
__000001 not found
r(111);

. tab mynewvar mynewvar2, missing

           |       mynewvar2
  mynewvar |         0          1 |     Total
-----------+----------------------+----------
         1 |         2     10,349 |    10,351 
-----------+----------------------+----------
     Total |         2     10,349 |    10,351

The Stata 18 help files say:

markvar(newvar) generates an indicator variable that identifies the observations used in the tabulation

.

Yes, the table and table twoway option markvar(newvar) does what the help files say but why does it also create an error? It is possibly my user error (despite being a long-time user-programmer),

[Nick Cox]

I guess that Jeff Pitblado (StataCorp) may pick this up but in any case I would go straight to StataCorp technical support for this.

[Jeff Pitblado (StataCorp)]

Here I am.
Anders found bug in table with option markvar().

The last line in the modern table code is calling count with the temporary touse variable that was (just prior) renamed to the specified markvar(). We hope to have the fix out to you soon, but in the mean time you can use capture noisily prefix in your do-file to keep Stata from exiting it early; the only thing you are losing here is the scalar result r(N).