Rewrite download_changesets.sh

[grischard]

• Use curl instead of wget to fetch the list of changesets.
• Use mktemp instead of a hardcoded filename for the temporary file.
• Allow the output directory to be specified via a command-line argument.
• Add logging with detailed messages, including timestamps and errors.
• Implement retries with exponential backoff for failed downloads.
• Implement a dry-run mode to display the changesets that would be downloaded without actually downloading them.
• Allow users to specify a custom log file location.
• Output a summary report of the downloaded changesets, including total count and size.
• Implement a help message that displays the available options and their descriptions.

This hopefully makes download_changesets more flexible, robust, and user-friendly than the original.

[woodpeck]

Thank you. (Unsure how curl is better than wget but hey...) Note that exponential backoffs may not always solve a download issue - I have encountered changesets so big that they could not be downloaded no matter how long you waited. For these I had to resort to @pnorman's make_changeset.py that synthesizes a changeset from minutely diffs.

[AntonKhorev]

Why the "since" date defaults to 2013-11-01? Wouldn't you want some date before osm existed?

[SomeoneElseOSM]

Is this planned to cover #30 ("Added download_changesets_txt.sh") and #25 ("download_changesets_uid.sh")

[mmd-osm]

I have encountered changesets so big that they could not be downloaded no matter how long you waited.

@woodpeck : This is a bit unexpected. Do you happen to have a download URL to demonstrate the issue?

As always, if you encounter similar issues in the future, please report them in the repo implementing the respective endpoint (rather than some random place on the internet).

[woodpeck]

@pnorman do you by any chance remember why you had to build your python "changeset synthesizer"? regarding @mmd's question above. Possibly that was at a time when changesets still had 50k objects and were maybe not even served by cgimap?

[pnorman]

@pnorman do you by any chance remember why you had to build your python "changeset synthesizer"? regarding @mmd's question above. Possibly that was at a time when changesets still had 50k objects and were maybe not even served by cgimap?

Yes, it was before cgimap handled changesets, and they could time out for large ones. But I mainly built it because it was interesting.