当前位置: 首页>>代码示例>>Python>>正文


Python csv.writer函数代码示例

本文整理汇总了Python中backports.csv.writer函数的典型用法代码示例。如果您正苦于以下问题:Python writer函数的具体用法?Python writer怎么用?Python writer使用的例子?那么恭喜您, 这里精选的函数代码示例或许可以为您提供帮助。


在下文中一共展示了writer函数的15个代码示例,这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞,您的评价将有助于系统推荐出更棒的Python代码示例。

示例1: fetch

    def fetch(self):
      fd, tmp_file = tempfile.mkstemp()

      pip = PointInPolygon(self.polygon_id, 60)

      traffic_signs = []
      reader = json.loads(open(self.mapping, 'r').read())
      try:
        for row in reader:
          traffic_signs += row['object']
      except:
        self.logger.err(row)
        raise

      with open(tmp_file, 'w') as csvfile:
        writer = csv.writer(csvfile)
        writer.writerow(['accuracy', 'direction', 'image_key', 'first_seen_at', 'last_seen_at', 'value', 'X', 'Y'])

      slice = lambda A, n: [A[i:i+n] for i in range(0, len(A), n)]

      bboxes = pip.bboxes()

      start_time = (datetime.today() - timedelta(days=365*2)).isoformat()[0:10]
      b = 0
      for traffic_signs_ in slice(traffic_signs, 10):
        b = b + 1
        self.logger.log('Batch {0}/{1}: {2}'.format(b, round(len(traffic_signs) / 10 + 0.5), ','.join(traffic_signs_)))
        for bbox in bboxes:
          url = 'https://a.mapillary.com/v3/map_features?bbox={bbox}&client_id={client_id}&layers={layer}&per_page=1000&start_time={start_time}&values={values}'.format(bbox=','.join(map(str, bbox)), layer=self.layer, client_id='MEpmMTFQclBTUWlacjV6RTUxWWMtZzo5OTc2NjY2MmRiMDUwYmMw', start_time=start_time, values=','.join(traffic_signs_))
          print(url)
          with open(tmp_file, 'a') as csvfile:
            writer = csv.writer(csvfile)

            r = None
            page = 0
            while(url):
              page = page + 1
              self.logger.log("Page {0}".format(page))
              r = downloader.get(url)
              url = r.links['next']['url'] if 'next' in r.links else None

              features = r.json()['features']
              filtered = 0
              self.logger.log('{0} features fetched'.format(len(features)))
              for j in features:
                p = j['properties']
                image_key = p['detections'][0]['image_key']
                gc = j['geometry']['coordinates']
                row = [p['accuracy'], p['direction'] if 'direction' in p else None, image_key, p['first_seen_at'], p['last_seen_at'], p['value']] + gc
                if row[0] > 0.01 and pip.point_inside_polygon(gc[0], gc[1]):
                  writer.writerow(row)
                  filtered = filtered + 1
              self.logger.log('{0} keeped'.format(filtered))

      return tmp_file
开发者ID:tkas,项目名称:osmose-backend,代码行数:55,代码来源:Analyser_Merge_Mapillary.py

示例2: test_writerows

    def test_writerows(self):
        class BrokenFile:
            def write(self, buf):
                raise OSError
        writer = csv.writer(BrokenFile())
        self.assertRaises(OSError, writer.writerows, [['a']])

        with TemporaryFile("w+", newline='') as fileobj:
            writer = csv.writer(fileobj)
            self.assertRaises(TypeError, writer.writerows, None)
            writer.writerows([['a','b'],['c','d']])
            fileobj.seek(0)
            self.assertEqual(fileobj.read(), "a,b\r\nc,d\r\n")
开发者ID:gitter-badger,项目名称:backports.csv,代码行数:13,代码来源:tests.py

示例3: _write_test

 def _write_test(self, fields, expect, **kwargs):
     with TemporaryFile("w+", newline='') as fileobj:
         writer = csv.writer(fileobj, **kwargs)
         writer.writerow(fields)
         fileobj.seek(0)
         self.assertEqual(fileobj.read(),
                          expect + writer.dialect.lineterminator)
开发者ID:gitter-badger,项目名称:backports.csv,代码行数:7,代码来源:tests.py

示例4: export_to_csv

    def export_to_csv(self, result_list, export_filename="ACRCloud_ScanFile_Results.csv", export_dir="./"):
        try:
            results = []
            for item in result_list:
                filename = item["file"]
                timestamp = item["timestamp"]
                jsoninfo = item["result"]
                if "status" in jsoninfo and jsoninfo["status"]["code"] == 0:
                    row = self.parse_data(jsoninfo)
                    row = [filename, timestamp] + list(row)
                    results.append(row)

            results = sorted(results, key=lambda x:x[1])

            export_filepath = os.path.join(export_dir, export_filename)

            with codecs.open(export_filepath, 'w', 'utf-8-sig') as f:
                head_row = ['filename', 'timestamp',  'custom_files_title', 'custom_acrid', 'title', 'artists', 'album',
                        'acrid', 'played_duration', 'label', 'isrc', 'upc', 'dezzer', 'spotify', 'itunes', 'youtube']
                dw = csv.writer(f)
                dw.writerow(head_row)
                dw.writerows(results)
                if self.debug:
                    self.log.info("export_to_csv.Save Data to csv: {0}".format(export_filename))
        except Exception as e:
            self.log.error("[email protected]_to_csv", exc_info=True)
开发者ID:acrcloud,项目名称:acrcloud_scan_files_python,代码行数:26,代码来源:acrcloud_scan_files_libary.py

示例5: write_csv

def write_csv(filename, rows):
    with io.open(filename, 'w', newline='') as f:
        writer = csv.writer(f)
        writer.writerow(["latitude", "longitude", "datetime"])
        for row in rows:
            writer.writerow([row["latitude"], row["longitude"], row["datetime"]])
        f.close()
开发者ID:jonaqp,项目名称:Django_Celery_Ejemplo,代码行数:7,代码来源:tasks.py

示例6: writeUniqueResults

def writeUniqueResults(clustered_dupes, input_file, output_file):

    # Write our original data back out to a CSV with a new column called 
    # 'Cluster ID' which indicates which records refer to each other.

    logging.info('saving unique results to: %s' % output_file)

    cluster_membership = {}
    for cluster_id, (cluster, score) in enumerate(clustered_dupes):
        for record_id in cluster:
            cluster_membership[record_id] = cluster_id

    unique_record_id = cluster_id + 1

    writer = csv.writer(output_file)

    reader = csv.reader(StringIO(input_file))

    heading_row = next(reader)
    heading_row.insert(0, u'Cluster ID')
    writer.writerow(heading_row)

    seen_clusters = set()
    for row_id, row in enumerate(reader):
        if row_id in cluster_membership:
            cluster_id = cluster_membership[row_id]
            if cluster_id not in seen_clusters:
                row.insert(0, cluster_id)
                writer.writerow(row)
                seen_clusters.add(cluster_id)
        else:
            cluster_id = unique_record_id
            unique_record_id += 1
            row.insert(0, cluster_id)
            writer.writerow(row)
开发者ID:rlugojr,项目名称:csvdedupe,代码行数:35,代码来源:csvhelpers.py

示例7: get

    def get(self, request, *args, **kwargs):
        object_list = self.get_queryset()[:2000]

        # Do reasonable ACL check for global
        acl_obj = self.translation or self.component or self.project
        if not acl_obj:
            for change in object_list:
                if change.component:
                    acl_obj = change.component
                    break

        if not request.user.has_perm('change.download', acl_obj):
            raise PermissionDenied()

        # Always output in english
        activate('en')

        response = HttpResponse(content_type='text/csv; charset=utf-8')
        response['Content-Disposition'] = 'attachment; filename=changes.csv'

        writer = csv.writer(response)

        # Add header
        writer.writerow(('timestamp', 'action', 'user', 'url', 'target'))

        for change in object_list:
            writer.writerow((
                change.timestamp.isoformat(),
                change.get_action_display(),
                change.user.username if change.user else '',
                get_site_url(change.get_absolute_url()),
                change.target,
            ))

        return response
开发者ID:dekoza,项目名称:weblate,代码行数:35,代码来源:changes.py

示例8: test_roundtrip_escaped_unquoted_newlines

 def test_roundtrip_escaped_unquoted_newlines(self):
     with TemporaryFile("w+", newline="") as fileobj:
         writer = csv.writer(fileobj, quoting=csv.QUOTE_NONE, escapechar="\\")
         rows = [["a\nb", "b"], ["c", "x\r\nd"]]
         writer.writerows(rows)
         fileobj.seek(0)
         for i, row in enumerate(csv.reader(fileobj, quoting=csv.QUOTE_NONE, escapechar="\\")):
             self.assertEqual(row, rows[i])
开发者ID:ryanhiebert,项目名称:backports.csv,代码行数:8,代码来源:tests.py

示例9: compare_dialect_123

    def compare_dialect_123(self, expected, *writeargs, **kwwriteargs):

        with TemporaryFile("w+", newline="", encoding="utf-8") as fileobj:

            writer = csv.writer(fileobj, *writeargs, **kwwriteargs)
            writer.writerow([1, 2, 3])
            fileobj.seek(0)
            self.assertEqual(fileobj.read(), expected)
开发者ID:ryanhiebert,项目名称:backports.csv,代码行数:8,代码来源:tests.py

示例10: test_unicode_write

 def test_unicode_write(self):
     import io
     with TemporaryFile("w+", newline='', encoding="utf-8") as fileobj:
         writer = csv.writer(fileobj)
         writer.writerow(self.names)
         expected = ",".join(self.names)+"\r\n"
         fileobj.seek(0)
         self.assertEqual(fileobj.read(), expected)
开发者ID:gitter-badger,项目名称:backports.csv,代码行数:8,代码来源:tests.py

示例11: test_roundtrip_quoteed_newlines

 def test_roundtrip_quoteed_newlines(self):
     with TemporaryFile("w+", newline="") as fileobj:
         writer = csv.writer(fileobj)
         self.assertRaises(TypeError, writer.writerows, None)
         rows = [["a\nb", "b"], ["c", "x\r\nd"]]
         writer.writerows(rows)
         fileobj.seek(0)
         for i, row in enumerate(csv.reader(fileobj)):
             self.assertEqual(row, rows[i])
开发者ID:ryanhiebert,项目名称:backports.csv,代码行数:9,代码来源:tests.py

示例12: test_quote_nonnumeric_decimal

    def test_quote_nonnumeric_decimal(self):
        """Decimals should not be quoted with non-numeric quoting."""
        import decimal

        with TemporaryFile("w+", newline="", encoding="utf-8") as fileobj:
            writer = csv.writer(fileobj, quoting=csv.QUOTE_NONNUMERIC)
            writer.writerow([10, 10.0, decimal.Decimal("10.0"), "10.0"])
            expected = '10,10.0,10.0,"10.0"\r\n'
            fileobj.seek(0)
            self.assertEqual(fileobj.read(), expected)
开发者ID:ryanhiebert,项目名称:backports.csv,代码行数:10,代码来源:tests.py

示例13: test_char_write

    def test_char_write(self):
        import array, string
        a = array.array(str('u'), text_type(string.ascii_letters))

        with TemporaryFile("w+", newline='') as fileobj:
            writer = csv.writer(fileobj, dialect="excel")
            writer.writerow(a)
            expected = ",".join(a)+"\r\n"
            fileobj.seek(0)
            self.assertEqual(fileobj.read(), expected)
开发者ID:gitter-badger,项目名称:backports.csv,代码行数:10,代码来源:tests.py

示例14: test_float_write

 def test_float_write(self):
     import array
     contents = [(20-i)*0.1 for i in range(20)]
     a = array.array(str('f'), contents)
     with TemporaryFile("w+", newline='') as fileobj:
         writer = csv.writer(fileobj, dialect="excel")
         writer.writerow(a)
         expected = ",".join([str(i) for i in a])+"\r\n"
         fileobj.seek(0)
         self.assertEqual(fileobj.read(), expected)
开发者ID:gitter-badger,项目名称:backports.csv,代码行数:10,代码来源:tests.py

示例15: get

    def get(self):
        f = io.StringIO()
        writer = csv.writer(f)

        headers = [
            'User ID',
            'Username',
            'First Name',
            'Last Name',
            'Email',
            'Telephone',
            'Enabled',
            'Admin',
            'Last Login',
            'Last Active',
            'Cohorts',
            'Hospitals',
            'Roles',
        ]
        writer.writerow(headers)

        def get_groups(user, group_type):
            """Comma-separated list of groups."""

            groups = [x.name for x in user.groups if x.type == group_type]
            groups = sorted(groups)
            groups = uniq(groups)
            return ', '.join(groups)

        def get_roles(user):
            """Comma-separated list of roles."""
            roles = [gu.role.name for gu in user.group_users]
            return ', '.join(sorted(set(roles)))

        users = list_users()
        for user in users:
            output = []
            output.append(user.id)
            output.append(user.username)
            output.append(user.first_name)
            output.append(user.last_name)
            output.append(user.email)
            output.append(user.telephone_number)
            output.append(user.is_enabled)
            output.append(user.is_admin)
            output.append(user.last_login_date)
            output.append(user.last_active_date)
            output.append(get_groups(user, GROUP_TYPE.COHORT))
            output.append(get_groups(user, GROUP_TYPE.HOSPITAL))
            output.append(get_roles(user))

            writer.writerow(output)

        return Response(f.getvalue(), content_type='text/csv')
开发者ID:renalreg,项目名称:radar,代码行数:54,代码来源:users.py


注:本文中的backports.csv.writer函数示例由纯净天空整理自Github/MSDocs等开源代码及文档管理平台,相关代码片段筛选自各路编程大神贡献的开源项目,源码版权归原作者所有,传播和使用请参考对应项目的License;未经允许,请勿转载。