worldbank/ietoolkit

iedorep review Sep 2022 - Errors

luisesanmartin opened this issue · 2 comments

I tried running the most recent dev version of iedorep with real reproducibility packages. Found a few errors which I'm listing below:

  • Do-files with comments that start with the characters // cause an error, as in
//	Set folder paths -----------------------------------------------------------
  • This line in an input do-file also caused an error
	global do           		"$user_git/03 – Data/07_Covid19 HA Data/Replication Package/dofiles"

When I changed global do to global do33 the error didn't appear anymore.

The error shown in both cases was this:
image

  • I also found this other error, but I don't know exactly what originates it
    image
    The first time it showed, it appeared after I had run another do-file which left a dataset loaded in memory. After typing clear and trying again, it didn't show up. Then I tried generating a dummy dataset by typing set obs 12 and gen n = _n, tried iedorep master.do again, and the same error showed. After typing clear again it didn't show up. I tried replicating it with another Stata session by typing again set obs 12 and gen n = _n, and running iedorep on another input do-file, but it only seems to show when I try it with this particular master do-file that I can share with you

Do-files with comments that start with the characters // cause an error
This line in an input do-file also caused an error

Neither of these are replicated generically in tests of the form added in 1ac8320:

// TEST COMMENT

global do "nothing"

so there must be something else going on. The post error you posted here usually stems from memory management that is interfering with the temp files I use to catch data states and record changes; that's why I already catch clear all and ieboilstart. But if this is being done in a sub-file it wouldn't get caught and will cause failure. For frames, I don't know the functionality well enough to predict what collision errors would happen but I have some ideas. I'll have to see this complete file to resolve these.

Solved in c4262e5. I was missing an instance of ieboilstart due to a whitespace typo (my fault) and then needed to add estimates clear to its replacement to avoid memory issues.