Using csv.DictWriter to output an in-memory gzipped csv file?

Question

I want to use a DictWriter from Python's csv module to generate a .csv file that's compressed using GZip. I need to do this all in-memory, so utilizing local files is out of the question.

However, I'm having trouble dealing with each module's type requirements in Python 3. Assuming that I got the general structure correctly, I can't make both modules work together because DictWriter needs to write to a io.StringIO buffer, while GZip needs a io.BytesIO object.

So, when I try to do:

buffer = io.BytesIO()
compressed = gzip.GzipFile(fileobj=buffer, mode='wb')
dict_writer = csv.DictWriter(buffer, ["a", "b"], extrasaction="ignore")

I get:

TypeError: a bytes-like object is required, not 'str'

And trying to use io.StringIO with GZip doesn't work either. How can I go about this?

Answer 1

You can use io.TextIOWrapper to seamlessly transform a text stream into a binary one:

import io
import gzip
import csv
buffer = io.BytesIO()
with gzip.GzipFile(fileobj=buffer, mode='wb') as compressed:
    with io.TextIOWrapper(compressed, encoding='utf-8') as wrapper:
        dict_writer = csv.DictWriter(wrapper, ["a", "b"], extrasaction="ignore")
        dict_writer.writeheader()
        dict_writer.writerows([{'a': 1, 'b': 2}, {'a': 4, 'b': 3}])
print(buffer.getvalue()) # dump the compressed binary data
buffer.seek(0)
dict_reader = csv.DictReader(io.TextIOWrapper(gzip.GzipFile(fileobj=buffer, mode='rb'), encoding='utf-8'))
print(list(dict_reader)) # see if uncompressing the compressed data gets us back what we wrote

This outputs:

b'\x1f\x8b\x08\x00\x9c6[\\\x02\xffJ\xd4I\xe2\xe5\xe52\xd41\x02\x92&:\xc6@\x12\x00\x00\x00\xff\xff\x03\x00\x85k\xa2\x9e\x12\x00\x00\x00'
[OrderedDict([('a', '1'), ('b', '2')]), OrderedDict([('a', '4'), ('b', '3')])]

Answer 2

A roundabout way would be to write it to a io.StringIO object first and then convert the content back to io.BytesIO:

s = io.StringIO()
b = io.BytesIO()

dict_writer = csv.DictWriter(s, ["a", "b"], extrasaction="ignore")

... # complete your write operations ...

s.seek(0)  # reset cursor to the beginning of the StringIO stream
b.write(s.read().encode('utf-8')) # or an encoding of your choice

compressed = gzip.GzipFile(fileobj=b, mode='wb')

... 

s.close()   # Remember to close your streams!
b.close()

Though as @wwii's comment suggest, depending on the size of your data, perhaps it's more worthwhile to write your own csv in bytes instead.