#!/usr/bin/env python # -*- coding: utf-8 -*- import glob from nose.tools import * from utilities import execution_path import os, mapnik2 def setup(): # All of the paths used are relative, if we run the tests # from another directory we need to chdir() os.chdir(execution_path('.')) if 'csv' in mapnik2.DatasourceCache.instance().plugin_names(): def get_csv_ds(filename): return mapnik2.Datasource(type='csv',file=os.path.join('../data/csv/',filename),quiet=True) def test_broken_files(visual=False): broken = glob.glob("../data/csv/fails/*.*") broken.extend(glob.glob("../data/csv/warns/*.*")) # Add a filename that doesn't exist broken.append("../data/csv/fails/does_not_exist.csv") for csv in broken: throws = False if visual: try: ds = mapnik2.Datasource(type='csv',file=csv,strict=True,quiet=True) print '\x1b[33mfailed\x1b[0m',csv except Exception: print '\x1b[1;32m✓ \x1b[0m', csv def test_good_files(visual=False): good_files = glob.glob("../data/csv/*.*") good_files.extend(glob.glob("../data/csv/warns/*.*")) for csv in good_files: if visual: try: ds = mapnik2.Datasource(type='csv',file=csv,quiet=True) print '\x1b[1;32m✓ \x1b[0m', csv except Exception: print '\x1b[33mfailed\x1b[0m',csv def test_type_detection(**kwargs): ds = get_csv_ds('nypd.csv') eq_(ds.fields(),['Precinct','Phone','Address','City','geo_longitude','geo_latitude','geo_accuracy']) eq_(ds.field_types(),['str','str','str','str','float','float','str']) feat = ds.featureset().next() attr = {'City': u'New York, NY', 'geo_accuracy': u'house', 'Phone': u'(212) 334-0711', 'Address': u'19 Elizabeth Street', 'Precinct': u'5th Precinct', 'geo_longitude': -70, 'geo_latitude': 40} eq_(feat.attributes,attr) eq_(len(ds.all_features()),2) def test_skipping_blank_rows(**kwargs): ds = get_csv_ds('blank_rows.csv') eq_(ds.fields(),['x','y','name']) eq_(ds.field_types(),['int','int','str']) eq_(len(ds.all_features()),2) def test_empty_rows(**kwargs): ds = get_csv_ds('empty_rows.csv') eq_(len(ds.fields()),10) eq_(len(ds.field_types()),10) eq_(ds.fields(),['x', 'y', 'text', 'date', 'integer', 'boolean', 'float', 'time', 'datetime', 'empty_column']) eq_(ds.field_types(),['int', 'int', 'str', 'str', 'int', 'bool', 'float', 'str', 'str', 'str']) fs = ds.featureset() feat = fs.next() attr = {'x': 0, 'empty_column': None, 'text': u'a b', 'float': 1.0, 'datetime': u'1971-01-01T04:14:00', 'y': 0, 'boolean': True, 'time': u'04:14:00', 'date': u'1971-01-01', 'integer': 40} eq_(feat.attributes,attr) while feat: eq_(len(feat),10) eq_(feat['empty_column'],None) feat = fs.next() def test_slashes(**kwargs): ds = get_csv_ds('has_attributes_with_slashes.csv') eq_(len(ds.fields()),3) fs = ds.all_features() eq_(fs[0].attributes,{'x':0,'y':0,'name':u'a/a'}) eq_(fs[1].attributes,{'x':1,'y':4,'name':u'b/b'}) eq_(fs[2].attributes,{'x':10,'y':2.5,'name':u'c/c'}) def test_wkt_field(**kwargs): ds = get_csv_ds('wkt.csv') eq_(len(ds.fields()),2) eq_(ds.fields(),['type','WKT']) eq_(ds.field_types(),['str','str']) fs = ds.all_features() #import pdb;pdb.set_trace() eq_(len(fs[0].geometries()),1) eq_(fs[0].geometries()[0].type(),mapnik2.GeometryType.Point) eq_(len(fs[1].geometries()),1) eq_(fs[1].geometries()[0].type(),mapnik2.GeometryType.LineString) eq_(len(fs[2].geometries()),1) eq_(fs[2].geometries()[0].type(),mapnik2.GeometryType.Polygon) eq_(len(fs[3].geometries()),1) # one geometry, two parts eq_(fs[3].geometries()[0].type(),mapnik2.GeometryType.Polygon) # tests assuming we want to flatten geometries eq_(len(fs[4].geometries()),4) eq_(fs[4].geometries()[0].type(),mapnik2.GeometryType.Point) eq_(len(fs[5].geometries()),2) eq_(fs[5].geometries()[0].type(),mapnik2.GeometryType.LineString) eq_(len(fs[6].geometries()),2) eq_(fs[6].geometries()[0].type(),mapnik2.GeometryType.Polygon) eq_(len(fs[7].geometries()),2) eq_(fs[7].geometries()[0].type(),mapnik2.GeometryType.Polygon) # ideally we should not have to: # https://github.com/mapnik/mapnik/issues?labels=multigeom+robustness&sort=created&direction=desc&state=open&page=1 if __name__ == "__main__": setup() [eval(run)(visual=True) for run in dir() if 'test_' in run]