Permalink
Commits on Jul 11, 2016
Commits on Jun 8, 2016
  1. Bump version for release.

    committed Jun 8, 2016
  2. Values larger than 64kB (uint16_t) could not be handled by the code. …

    …The only place it was possible for this to happen was in changeset comments, which can apparently be quite large. This adds the ability to extend the length if necessary up to 64 bits, which should be sufficient.
    committed Jun 8, 2016
  3. Handle case where author of a comment isn't a public user. This shoul…

    …dn't happen on the real dumps, as public users can no longer make changes.
    committed Jun 8, 2016
Commits on Jan 15, 2016
  1. Bump version for release.

    committed Jan 15, 2016
  2. Include vendored Boost header, as the gzip filter in boost versions <…

    …1.60 has a bug which doesn't write the gzip header to empty files, which makes it create invalid gzip files.
    committed Jan 15, 2016
Commits on Jan 14, 2016
Commits on Jul 10, 2015
  1. Hack to prevent oversized relations.

    PBF format requires that the uncompressed size of a block is no
    larger than 32MB. Calculating the size of a block has a cost, so
    it is only done periodically (for relations, every 2000 elements).
    
    The issue arises because 2000 relations manage to be larger than
    24MB (the difference between the cut-off and the max size). The
    quick hack fix for this is to reduce the periodic check down to
    every 500 relations.
    
    This is obviously not a real fix; for that, we'll need to be able
    to calculate the size of each element before adding it to the
    block, including the size of any new strings added to the string
    table. And, moreover, do that in an efficient manner.
    
    However, this fix should be enough to allow us to get the next
    planet out.
    committed Jul 10, 2015
Commits on Jun 11, 2015
  1. Saner default for handling existing disk DBs.

    Previously, the default behaviour was to try to resume processing
    using any existing data that was present. This had the unintended
    side effect of silently using old data if it wasn't cleaned up.
    
    I'm sure this being the default made sense at the time, but it
    doesn't any more and it confused me greatly. Therefore, the new
    default is to remove any existing data and start from scratch. The
    old behaviour is available using the `--resume` flag.
    committed Jun 11, 2015
Commits on Feb 22, 2015
  1. Merge pull request #11 from zerebubuth/remove-leveldb

    Removed all references to LevelDB
    committed Feb 22, 2015
  2. Removed all references to LevelDB - it has not been the recommended d…

    …isk database format for a while now.
    committed Feb 22, 2015
  3. Merge pull request #8 from pnorman/patch-1

    Remove leveldb references
    committed Feb 22, 2015
  4. Merge pull request #7 from pnorman/no_bash

    use sh instead of bash for autogen
    committed Feb 22, 2015
  5. Made generator name configurable via the command line.

    The generator name appears in the output files (both XML and PBF)
    which makes it annoying to test, as a bumped version invalidates
    all the static test comparison files. To fix it, this adds an
    option to override the `PACKAGE_STRING` so that the tests can use
    a stable value. It is not recommended for other use, as having
    the current version in the dump file is useful for tracing back
    any errors or bug reports.
    committed Feb 22, 2015
Commits on Feb 21, 2015
  1. Merge pull request #10 from zerebubuth/changeset-comment-dumps

    Changeset comment dumps
    committed Feb 21, 2015
  2. Ignore test-related files.

    committed Feb 21, 2015
  3. Added test cases.

    committed Feb 21, 2015
  4. Initialise 'zero' entries properly.

    Previously, initial entries were initialized to zero - a value for
    element IDs which doesn't exist in the database. However, when a
    database is loaded using Osmosis, it does create some entries with
    ID=0. Therefore, it's better to just initialise to a negative
    number and ensure that negative elements are never processed.
    committed Feb 21, 2015
  5. Add support for changeset discussions in the dump.

    This adds support for an *additional* dump of changesets which
    includes discussions. This will be the new format and deprecates
    the previous format, but we'll allow a few months after the
    announcement for the previous format to be sunsetted.
    
    The format is the same as getting changesets with discussions from
    the XML API.
    committed Feb 21, 2015
Commits on Feb 17, 2015
  1. Use an array of status variables, rather than a single index for the …

    …finished thread as more than one thread can finish before the main thread wakes up.
    committed Feb 17, 2015
Commits on Dec 24, 2014
  1. Release v1.0.0.

    committed Dec 24, 2014
Commits on Mar 5, 2014
  1. Merge remote-tracking branch 'origin/master' into changesets

    Conflicts:
    	src/Makefile.am
    	src/planet-dump.cpp
    committed Mar 5, 2014
  2. Fix typo in XML writer.

    Many thanks to @pa5cal for pointing out, in issue #4, that the
    relation member type was being determined from the member's ID,
    rather than its actual type.
    
    This fixes #4 by using the `member_type` field correctly.
    committed Mar 5, 2014
Commits on Feb 20, 2014