How-to output CSV from Django should suggest using `StreamingHttpResponse`

[Simon Charette]

The ​Outputting CSV with Django how-to doesn't even mention StreamingHttpResponse even if ​it’s useful for generating large CSV files.

I suggest we replace the example with something along the following:

import csv
from StringIO import StringIO

from django.http import StreamingHttpResponse


def some_view(request):
    rows = (
        ['First row', 'Foo', 'Bar', 'Baz'],
        ['Second row', 'A', 'B', 'C', '"Testing"', "Here's a quote"]
    )

    # Define a generator to stream data directly to the client
    def stream():
        buffer_ = StringIO()
        writer = csv.writer(buffer_)
        for row in rows:
            writer.writerow(row)
            buffer_.seek(0)
            data = buffer_.read()
            buffer_.seek(0)
            buffer_.truncate()
            yield data

    # Create the streaming response  object with the appropriate CSV header.
    response = StreamingHttpResponse(stream(), content_type='text/csv')
    response['Content-Disposition'] = 'attachment; filename="somefilename.csv"'

    return response

[Daniele Procida]

Yes, and also ​https://docs.djangoproject.com/en/dev/ref/request-response/#django.http.StreamingHttpResponse should link to this, in the text "For instance, it’s useful for generating large CSV files"

[Marc Tamlyn]

I'm not convinced. I've output many a CSV file and never needed the streaming response to get performance. Whilst this is a useful addition to mention at this point in the docs, I don't think we should be recommending the more complex option.

[Aymeric Augustin]

The code example looks like C, not like Python... I don't want to see buffer_.seek(0) in our docs.

Streaming responses don't change much when you pull all the data in RAM, and if the data comes from a queryset, Django currently does that even if you use .iterator(). It seems much more interesting to me to optimize the database side than the HTTP response side.

[Simon Charette]

Thinking about it I must agree that without server-side cursor support (#16614) the tradeoff is not worth turning the simple example into a overly complex one.

I just thought it was odd that StreamingHttpResponse's documentation mentions that it’s useful for generating large CSV files but ​our provided tutorial doesn't even mention it.

What do you guys think of adding an admonition with no specific example to the how-to explaining StreamingHttpResponse might be useful in this case?

[Daniele Procida]

StreamingHttpResponse could still do with some example code in the docs, even if it doesn't replace the existing example.

[ANUBHAV JOSHI]

Any ideas regarding what type of example should be given in the docs for StreamingHttpResponse?

[Rigel Di Scala]

Hello, I would like to work on this ticket.

I think that some information on how to test a view that returns a StreamingHttpResponse() would be useful. The Django test Client actually returns an iterable response, and the .streaming_content property is an instance of <itertools.imap>. You would then need to concatenate it into a string in order to test it, as you would do with the standard HttpResponse.

[Rigel Di Scala]

I was thinking of something along these lines:

import csv

from django.http import StreamingHttpResponse


class Echo(object):
    def write(self, value):
        return value


def some_streaming_view(request):
    rows = (["Row {0}".format(idx), str(idx)] for idx in xrange(100))
    buffer_ = Echo()
    writer = csv.writer(buffer_)
    response = StreamingHttpResponse((writer.writerow(row) for row in rows),
                                     content_type="text/csv")
    response['Content-Disposition'] = 'attachment; filename="somefilename.csv"'
    return response

I have tested it with curl, a simple test case with the Django test client, and a regular browser.

[Rigel Di Scala]

You can also test this with an infinite series, such as the classic Fibonacci function, if you replace the range generator with something like:

I tested this and the memory use did not increase significantly even after streaming over a gigabyte of data for a single request.

[Daniele Procida]

The example above looks good to me. Please do submit a pull request - thanks.

[Rigel Di Scala]

I have opened a pull request here:

​https://github.com/django/django/pull/2358

I am using a slight variation of the above example, using Python 3 friendly code and some additional comments, as suggested by bmispelon.

[Rigel Di Scala]

Resubmitted a new pull request: ​https://github.com/django/django/pull/2397

[Tim Graham <timograham@…>]

In fad47367bf622635b4cf931db72310cce41cebb4:

Fixed #21179 -- Added a StreamingHttpResponse example for CSV files.

Thanks charettes for the suggestion.