Failures cause duplicates

[wurli]

If the process fails after tweets are posted, the whole process will fail, meaning tweets will be posted again next time. If this fails again then this can result in a loop of updates getting repeatedly posted.

This happened when a tonne of changes were made to the dplyr README before a release (tidying up). I think the volume of tweets this generated led to a failure, and has done so on every subsequent run (when the full bunch of updates is attempted to be posted).

Ways to fix this:

• Add some tryCatch() stuff to make it almost impossible for things to fail on the R-side if posting has succeeded. This should be done anyway, because if the exact text of a tweet is attempted to be posted a second time, this will (seemingly) always fail.
• Test if the text of a tweet has already been posted (using {rtweet}) before posting... More like avoiding failures in the first place
• Only post the first, say, 10 tweets generated in a run. This would at least mean the next bunch would be deferred until the following run.