Field names mixed up writing via "KML" driver

Expected behavior and actual behavior.

Attempting to save a GeoPandas dataframe using:

        with fiona.Env():
            gdf.to_file(kml_path, driver='KML', NameField='name')

produces a KML file in which the value for <name> is taken from another field, and likewise <description>. If the file is written the hard way (using ogr) it is correct. Perhaps there are some kwargs that need to be passed along to correct this? In the example above NameField='name' was an unsuccessful attempt.

Correct KML using ogr:

<?xml version="1.0" encoding="utf-8" ?>
<kml xmlns="http://www.opengis.net/kml/2.2">
<Document id="root_doc">
<Schema name="walk" id="walk">
	<SimpleField name="id" type="string"></SimpleField>
	<SimpleField name="date" type="string"></SimpleField>
	<SimpleField name="distance" type="string"></SimpleField>
	<SimpleField name="speed" type="string"></SimpleField>
	<SimpleField name="time" type="string"></SimpleField>
	<SimpleField name="activity_name" type="string"></SimpleField>
</Schema>
<Folder><name>walk</name>
  <Placemark>
	<name>Walked 3.97 km on 22/1/17</name>
	<Style><LineStyle><color>ff0000ff</color></LineStyle><PolyStyle><fill>0</fill></PolyStyle></Style>
	<ExtendedData><SchemaData schemaUrl="#walk">
		<SimpleData name="id">1965703340</SimpleData>
		<SimpleData name="date">01/22/2017</SimpleData>
		<SimpleData name="distance">3.96978493824</SimpleData>
		<SimpleData name="speed">1.397811597971831</SimpleData>
		<SimpleData name="time">2840</SimpleData>
		<SimpleData name="activity_name">Walk</SimpleData>
	</SchemaData></ExtendedData>
      <LineString><coordinates>151.165393442,-33.7688339688 ... 151.16551338,-33.7687453898</coordinates></LineString>
  </Placemark> ...

Incorrect KML using Fiona (see Placemark name and description):

<?xml version="1.0" encoding="utf-8" ?>
<kml xmlns="http://www.opengis.net/kml/2.2">
<Document id="root_doc">
<Schema name="walk" id="walk">
	<SimpleField name="id" type="string"></SimpleField>
	<SimpleField name="date" type="string"></SimpleField>
	<SimpleField name="distance" type="float"></SimpleField>
	<SimpleField name="speed" type="float"></SimpleField>
	<SimpleField name="time" type="string"></SimpleField>
	<SimpleField name="activity_name" type="string"></SimpleField>
</Schema>
<Folder><name>walk</name>
  <Placemark>
	<name>2840</name>
	<description>01/22/2017</description>
	<Style><LineStyle><color>ff0000ff</color></LineStyle><PolyStyle><fill>0</fill></PolyStyle></Style>
	<ExtendedData><SchemaData schemaUrl="#walk">
		<SimpleData name="id">3.96978493824</SimpleData>
		<SimpleData name="date">Walked 3.97 km on 22/1/17</SimpleData>
		<SimpleData name="distance">1.39781159797183</SimpleData>
		<SimpleData name="speed">0</SimpleData>
	</SchemaData></ExtendedData>
      <LineString><coordinates>151.165393442,-33.7688339688 ... 151.16551338,-33.7687453898</coordinates></LineString>
  </Placemark> ...

Steps to reproduce the problem.

I don’t think there was anything special about the GeoDataFrame, but can provide a script and sample data if the answer is not trivial.

Operating system

Mac OS X 10.15.4

Fiona and GDAL version and provenance

Fiona 1.8.13.post1 loaded by PyCharm GDAL 2.4.0 loaded by PyCharm

[wheeled]

So https://github.com/Toblerity/Fiona/blob/master/fiona/ogrext.pyx#L1119-L1122 explains how the fields get mixed up. If I simulate the behaviour of the dict(zip()) it produces a mapping that agrees with the result:

coll_schema = ["id", "date", "distance", "name", "speed", "time", "activity_name"]
ogr_schema = ["Name", "Description", "id", "date", "distance", "name", "speed", "time", "activity_name"]
schema_mapping = dict(zip(coll_schema, ogr_schema))
print(schema_mapping)

{'id': 'Name', 'date': 'Description', 'distance': 'id', 'name': 'date', 'speed': 'distance', 'time': 'name', 'activity_name': 'speed'}

[note that this process remaps time to name and then later on the ogr driver remaps name to Name]

I understand that you are not necessarily keen to support KML, but if this issue also affects some other drivers would it be possible to classify it as a bug?

I imagine the fix would be to check that the collection.schema['properties'].keys() and ogr_schema['properties'].keys() were the same length prior to zipping them. If all of the offending ogr drivers add their special fields at the beginning of the schema, this might work as a fix:

        ogr_schema = self.get_schema()['properties'].keys()
        coll_schema = collection.schema['properties'].keys()
        offset = len(ogr_schema) - len(coll_schema)
        self._schema_mapping = dict(zip(
            coll_schema,
            ogr_schema[offset:] ))

This assumes that the driver will look after its own Name and Description fields.

I apologise again that I am not set up to be able to test it.

[rbuffat]

The real issue is, that the KML driver automatically adds fields: https://github.com/OSGeo/gdal/blob/master/gdal/ogr/ogrsf_frmts/kml/ogrkmllayer.cpp#L116-L120

https://github.com/Toblerity/Fiona/blob/master/fiona/ogrext.pyx#L1119-L1122 assumes that only the fields in the schema provided to Fiona exist. The same issue is also present for the DXF, GPX, GPSTrackMacker and DGN driver.

Tested by adding:

diff --git a/fiona/ogrext.pyx b/fiona/ogrext.pyx
index a0b2a6c..c8163ca 100644
--- a/fiona/ogrext.pyx
+++ b/fiona/ogrext.pyx
@@ -1083,6 +1083,12 @@ cdef class WritingSession(Session):
 
             encoding = self._get_internal_encoding()
 
+            _schema = self.get_schema()
+            log.debug("before schema: {}".format(_schema))
+
+            if len(_schema['properties']) > 0:
+                raise ValueError("Driver {} has non empty schema: {}".format(self.collection.driver, _schema))
+
             for key, value in collection.schema['properties'].items():
 
                 log.debug("Begin creating field: %r value: %r", key, value)