[Talk-us] Another day, another bad import

[Mike N]

On 5/1/2011 6:19 PM, Ben Supnik wrote:
> If/when you make contact with the importer, it would be useful to know
> what went wrong exactly, e.g. did an automated part of the conversion
> fail, what was the last phase of the import the user checked, did the
> import fail on upload, etc.
>
> Given the discussion of NHD imports last week and my attempts to make
> the NHD data more manageable, it would be useful to know where the
> failure points are.

   Since JOSM was used in this case, I already know one problem from 
experience - if you upload from JOSM and the upload fails, you end up 
with a partial upload.   Recovery is ugly - a full revert of the 
changeset or a manual match against the changeset to figure out what was 
uploaded.

   Even the better bulk upload scripts that create a checkpoint file 
every N data points have a problem with the last chunk because they may 
think the upload failed, but it actually succeeded.   That requires a 
manual fix, or perhaps an automated version could query the last 
changeset to figure out exactly what was uploaded, and continue from there.