Non-ascii field names

[Vanuan]

In Python 2 variable names are limited by ascii character set.
CSV Scrapy exporter uses Item's field names as an csv title row.
I want to use unicode characters for csv title row.
Can I do that without writing my own exporter?

[ghost]

I think, the easiest way is - I often see this as the easiest way, so maybe I am wrong - to extend the existing class like this (untested, for reference):

class UnicodeCsvItemExporter(CsvItemExporter):
    def __init__(self, file, include_headers_line=True, join_multivalued=',', headers={}, **kwargs):
        super(UnicodeCsvItemExporter, self).__init__(file, include_headers_line, join_multivalued, **kwargs)

        self.header_mapping = headers

    def _write_headers_and_set_fields_to_export(self, item):
        if self.include_headers_line:
            if not self.fields_to_export:
                self.fields_to_export = item.fields.keys()
            # Maybe list(...) is required here to give writerow a list and not a lazy iterable
            self.csv_writer.writerow(map(lambda x: self.header_mapping[x], self.fields_to_export))

Then, you'd have to add your exporter to the settings:

FEED_FORMAT = 'csv_unicode'
FEED_EXPORTERS = {
    'csv_unicode': '.....',
}

Edit:
mmhh... or maybe not. You will not have access to the headers parameter through the string settings... If you find a solution for this, then it will be fine. Otherwise, it's not a solution.
My idea here would be to add another setting in settings.py called CSV_UNICODE_HEADERS, but it seems exporters are one of the few places not having a reference of settings.

I am now really wondering, where scrapy generates an instance of the selected exporter, because each class has a totally different constructor signature.

[nramirezuy]

@Vanuan Extend this method to pull the header name from the item.Field.