[Imports] Optimal ChangeSet size
grand.edgemaster at gmail.com
Sat Aug 1 23:43:49 BST 2009
I'll just update this thread with the news that in this past week I've
been hacking on the bulk_upload.py code, I made several changes,
including splitting apart the concepts of a Changeset and a Diffset
More recently, I hope I've made the code more accessible to other
developers who want to build on top of the bulk_import codebase, such
as I wished to for the NaPTAN uploader -
2009/7/22 Thomas Wood <grand.edgemaster at gmail.com>:
> Be aware that there is a difference between changeset size and diff upload size.
> bulk_upload sets the changeset size incorrectly at 1000, it means to
> set the diff upload size to that.
> The reasoning for this is that when diff uploads were first
> introduced, the server had some issues with the number of features it
> could process. I'm not sure if this has changed since.
> 2009/7/22 Mike N. <niceman at att.net>:
>> What is the optimal size of a changeset, and what are the tradeoffs in
>> changing the size? From what I understand, the API limit is 50000 , but it
>> would be good to keep below this so that later revisions to the changeset do
>> not push it over.
>> bulk_upload.php sets its limit at 1000 , which seems a bit small when
>> uploading lots of data - 50,000 nodes will create 50 changesets. Is
>> there any disadvantage to many changesets? Nodes are more likely to be in
>> different changesets than the ways.
>> Increasing the size might result in more lost upload time in case of a
>> network error - a larger changeset upload session will be lost. On the
>> other hand, if I increase the limit to 5000 or 10000 , does this slow down
>> future map edits by requiring fetching larger amounts of data from
>> changesets? (I'm not sure how that works).
>> Imports mailing list
>> Imports at openstreetmap.org
> Thomas Wood
More information about the Imports