* ENH: get SRID from postgis read
As currently implemented, this is relying on string parsing from the sql
string passed in by the user. An API change could achieve this without
the string parsing. Or something else I'm not seeing?
* Fix polygons -> geometries in comment
* Remove brittle SRID postgis extraction
Change previous test to xfail, add test of "select geom as the_geom".
* Use GEOSGetSRID to return SRID from postgis
Per @jorisvandenbossche suggestion. Also simplifies wkb loading using
the shapely.skb.loads argument hex=True. Add the argument hex_encoded to
read_postgis so that non hex geometries might also be read.
* Un-xfail test for SRID read
With SRID read from geodatabase support, can now set CRS automatically
so that user does not need to.
* Remove check for unique SRID in sql read
Assume the CRS for frame from SRID in first geometry.
* Add test of sqlite database read
* Add type check in sql geometry read
I think this both addresses python2 failures for sqlite and WKBElement
reads.
* Rename read_postgis to read_sql
Create wrappers read_postgis and read_sqlite
* Add read_sqlite to __init__
* Add test to create sqlite database for testing
Rename the function for postgis from create_db to create_postgis. This
permits not shipping the sqlite table as part of the datasets.
* Amend read_sql docs per Joris's comments
* Remove nybb.sqlite dataset
* Clean up sqlite db creation/execution
* Remove explicit sqlite support
* Remove read_sqlite command from __init__ too
* Fix sqlite read tests
* Fix failures in merge conflict resolution
I noticed some differences between the docs and the source code. By default scheme is ``None`` and ``equal_interval`` mustn't end with an s otherwise an exception is raised. I also changed the link because the current URL resolves to a 404 error.
Not doing so resulted in incorrect dtypes when the
result was empty since numpy could not automatically
determine the dtype. This then had the effect of
indexing dropping columns unexpectedly.
Closes#685
* Fix for when read_file bbox crs does not match target file crs
* added tests for read_file with GeoSeries/GeoDataFrame boundary
* Group related tests in test_io
* Make adjustment to read_file bounding box doc
* BUG: Save z dimension in file writes
In so doing, add the has_z attribute to series. The check in determining
the geometry type is pretty simplistic. Given how fiona is thinking
about geometry types now (https://github.com/Toblerity/Fiona/issues/465)
I didn't see a clearly better way.
* Add test for a 3d polygon, too
* TST: Add has_z test
* TST: Mix 2D/3D polygons in file writes
Removed the pandas assert_frame_equal test because it will no longer
pass anywhere due to the 2D/3D handling, which doesn't look like
something there's any way around without changes to fiona (or possibly
GDAL?).
* Remove descartes from install requirements
The descartes import in plotting is hidden, so it is not a hard
dependency, and should not be required for installation.
* Update install reqs in README.md
* Raise ValueError using to_file with bool column
Fiona doesn't support writing to files with boolean type. There was some
discussion of converting to int and writing to file anyway, but at a
minimum seems useful to raise a clearer message about the problem.
* Improve warning with to_file on bool column
Can optionally set the size of points in a scatter plot of point
geometries to be specified by the value of a column in the frame. Note
that it does not perform any normalization at present, so very small or
very large values in the column specified lead to less than desirable
results. Also no checks on user, so if a user specifies a column with a
non numeric type, it does throw a slightly less than clear error.
* fixed nybb
* rearrange chloropleths example
* added badge to MyBinder
* extra formatting on README
* add extra info in chrolopleths.ipynb
* Update choropleths example
* BUG: sjoin modifies original frame (#525)
Original pandas table's index names are no
longer changed by sjoin() call. Includes
some unit-tests. Fixes issue #525
* New code to fix issue that is compatible with pandas 0.16.2 - 0.20.3
* applied suggested changes to test_sjoin.py
As of 0.20.0, pandas .ix is deprecated. This switches all such instances
to the label based .loc indexer. (In searching for usage, I also noticed
the CHANGELOG contained an error in describing the introduction of the
.cx indexer.)
All current overlay tests have indices for which loc/iloc give the same
result. Since ix will be deprecated, add a test to distinguish behavior
between the two.
Joris Van den Bossche