Handling of out of range values

[Sergiy Radyakin]

Dear All,

I've just found that the following code does not produce an error, while I anticipated it should:

clear
generate x="abc"
display x[1]
display x[1000]
display x[-1000]

I expected to see here:

observation numbers out of range
r(198);

Similar to what we are getting after

replace x=1 in 1000

Observed in Stata 17 (if this matters, but probably all versions of Stata).

Is there any way to restrict this behavior to result in error for when attempts are made to read from outside of the data range?

Thank you, Sergiy

[Nick Cox]

It is not always an error to refer directly to an observation not in the dataset. A common example is something like

bysort foo : gen whatever = bar[_n-1]

which for the first observation in each group refers to bar[0], which is not in the dataset, but Stata just evaluates this as missing.

This is a an example of the error you mention, as there are 74 observations in the auto dataset.

. sysuse auto, clear
(1978 automobile data)

. replace mpg = 100 in 100
observation numbers out of range
r(198);

I don't have a good story for the reasons why or why not, but the first is something I have done many times on purpose, where anything like the second is just a silly mistake.