Error in strdup

[Sergiy Radyakin]

Dear All,

I am trying to understand the exact reason the string duplication fails sometimes.
The documentation illustrates some possible uses:

Yet, when I run this (highlighted) example in Stata I am getting an error message:

For the moment, I've noticed that enclosing the string duplication in parenthesis helps and can be used as a workaround.
I am also not sure whether the order matters in Num*String or String*Num, since Num*String seems to work fine.

The above is from Stata MP 16.0 (Windows 64-bit), but the same was earlier observed in Stata 15.1.

Thank you for any comments on the above behavior,
Sergiy

[William Lisowski]

The problem is not in duplicating strings, but rather with the display command, which is notorious for trying - and sometimes failing - to identify expressions and replace them with their evaluated values. Here are three more examples along the lines of yours.

. display `""hello"*3"'
"hello"*3

. display `="hello"*3'
hellohellohello not found
r(111);

. display "`="hello"*3'"
hellohellohello

When you remove the display command from the process, you can see that the string duplication works as documented when the expression is parsed as intended.

. local s1 = 3*"hello"

. local s2 = "hello"*3

. macro list _s1 _s2
_s1:            hellohellohello
_s2:            hellohellohello

[Sergiy Radyakin]

William, thank you for your thoughts.

Regarding the examples, I don't have any problem with the 1st and the 3rd examples you provided.

. display `""hello"*3"'
"hello"*3

. display "`="hello"*3'"
hellohellohello

I might agree to #2 as well, seeing how #3 works. Yet I don't get the difference between expressions 3*"A" and "A"*3 as seen by the display command. I have tried to find other examples, but can't, e.g. list command does not evaluate the expression 3*"A" and complains whichever way it is written, yet the if command is always evaluating the expression correctly, again, regardless which way it is written. I don't see why display is anything special here. For me it is a bug 🐛 in the display command then.

Thank you, Sergiy

[William Lisowski]

A close reading of [R] display tells us two of the many allowed display directives are

"double-quoted string"
[%fmt] [=] exp

Your problem can be solved by being explicit to display that you are presenting an expression to evaluate rather than a double-quoted string followed by something else.

. display "hello"*3
hello*3 invalid name
r(198);

. display = "hello"*3
hellohellohello

While I am not fond of the display command its problems arise from attempting to infer more than it can do successfully, rather than from a bug. It should have required the "=" before an expression, but that would have made the interactive use cumbersome to accommodate the very small percentage of edge cases. In your case, display treats "hello" as a string to display (note that it is not in red) and since the next tokens are not within a string, guesses that they form an expression to be evaluated, and throws an error in evaluating this expression, consistent with the error thrown by

. local s = * 3
*3 invalid name
r(198);

Statalist is rife with examples of display presenting unexpected results. My own preference as a programmer is to evaluate expressions into local macros and then present the results within a quoted string that includes the dereference of the local macro. That way, display isn't trying to infer what is wanted.