Possible bug in -reshape-

[Robert Ulbricht]

I am encountering an error when using -reshape- with 'favor(speed)' when subsetting j. Using 'favor(memory)' works as expected which leads me to suspect a bug? Running on Windows/MP18.5 (revision 2024-12-18).)

Silly example:

sysuse bplong.dta, clear
reshape wide bp, i(patient) j(when 2) favor(speed)

This throws the following error:

_reshape_df::increment_j():  3301  subscript invalid

[Jeff Pitblado (StataCorp)]

Thank you for providing a minimal example. We will look into this and hopefully fix it soon.

[Clyde Schechter]

This does seem to be a bug. I can replicate the findings reported in #1 on my setup. (Windows 10, Stata MP4 version 18.5)
Interestingly, if you list out all of the values of the -j()- variable, then the code does not throw an error. Similarly, if you first -keep if inlist(bp, 138, 141)- and run the code exactly as shown in #1, you also get no error. But it does seem that the -favor(speed)- version is not recognizing partial subset lists correctly.

Added: Crossed with #2.

[Jeff Pitblado (StataCorp)]

This has been fixed in the Stata 18 update on 26feb2025.

[Robert Ulbricht]

Thanks for providing a fix. I can confirm that the original bug has been fixed in that update.

However, there seems to have been a regression in Stata 19 where -reshape- now fails again under a similar set of circumstances.

I am attaching a full reproducer below.

/*
 * This is a minimal reproducer for the reshape bug on Stata 19.
 *
 * Setting `J' to 10 or less, or setting favor(memory) avoids the bug,
 * and produces the following dataset:
 *
 * . list  /* Result from `J' = 11 and favor(memory) */
 *
 *      +------------------------------------------------------+
 *      | i   x1   x2   x3   x4   x5   x6   x7   x8   x9   x10 |
 *      |------------------------------------------------------|
 *   1. | 1    0    0    0    0    0    0    0    0    0     0 |
 *      +------------------------------------------------------+
 *
 * Setting `J' to at least 11 AND setting favor(speed) causes a bug,
 * where the reshape fails with the following error:
 *
 * variable x2 not found
 *
 */

version 19
clear
 
local J 11
set obs `J'

gen i = 1
gen j = _n
gen x = 0

levelsof j if j < `J', local(values)
reshape wide x, i(i) j(j `values') favor(speed)

[Jeff Pitblado (StataCorp)]

Thank you for providing a minimal example. We hope to have this fixed in the next update to Stata (18 and 19...).