Write to specified file for dumpdata

[Gwildor Sok]

In response of ticket #22251, I'm opening this as a separate issue like requested. You can read the need for this option there, but basically it removes the current need and practice of redirecting the output of the command to a file, which is often done like this:

$ ./manage.py app1 app2 .. --format=json > dump.json

Removing the need to always do this would open up the possibility to add verbosity during the serialization (see #22251 and a new separate ticket for that).

Concrete proposal:
Add an option to specify a file name to write the output to (under the current working directory), and generate one automatically when one is not specific (e.g. "dump_20140312_1337".json/.xml").

[Aymeric Augustin]

Yes, that makes sense.

[ANUBHAV JOSHI]

I have already created patch for this. But one thing I want to ask is that whether we should keep default as stdout or as proposed in the ticket some dumpdata.json

[ANUBHAV JOSHI]

I have yet to add tests.
​https://github.com/coder9042/django/compare/ticket_22258?expand=1

Currently the option has been added but default is stdout.
I think that default should be as it is because there are times when data is not that big and people will like to see the result then and there itself, plus this way backward compatibility can be maintained.

[ANUBHAV JOSHI]

Added tests.
Created PR : ​https://github.com/django/django/pull/2465

[loic84]

Add an option to specify a file name to write the output to (under the current working directory)

Why under the current working directory? open() does the right thing by default, if given a relative path, it'll use the current working directory. The current working directory could very much be readonly, and requiring people to cd to target a specific directory is not practical.

From a quick look at the patch, I think it does the right thing.

[Gwildor Sok]

Replying to loic84:

Add an option to specify a file name to write the output to (under the current working directory)

Why under the current working directory? open() does the right thing by default, if given a relative path, it'll use the current working directory. The current working directory could very much be readonly, and requiring people to cd to target a specific directory is not practical.

From a quick look at the patch, I think it does the right thing.

I'm sorry, this was what I meant. So you can also do -o ../dumps/dump.json, or something alike. Patch indeed looks good on that matter.

About the default: I agree that defaulting to stdout is the way to go to maintain backwards compatibility. The main reason I opted for generating a filename and using that when not specific, is in the light of ticket #22258.

[ANUBHAV JOSHI]

Replying to anubhav9042:
Can you comment out your thoughts regarding #22258 on the ticket itself, I mean if you have something in mind

[ANUBHAV JOSHI]

Replying to Gwildor:

About the default: I agree that defaulting to stdout is the way to go to maintain backwards compatibility. The main reason I opted for generating a filename and using that when not specific, is in the light of ticket #22258.

Do you have any ideas in mind. If yes then please comment them out on #22258 itself. I want to know exactly what you are suggesting.
You want progress to be displayed.Fine, but it would be nice to have some elaborations.

Also as in #22251, by russellm:"we're limited by the progress reporting tools of the underlying serialisation frameworks. If you've got any ideas on how this can be achieved, I'd love to hear them."

[ANUBHAV JOSHI]

I think we can keep default as stdout and and for #22258, we can do the following:

• When -o is not used, let the data be printed to stdout, as the data is printing people can see it and I don't think any progress display is required there. Also people may not want to open a file everytime they perform dumpdata even for small databases. If its large they'll use -o.

• When -o will be used, then we can display progress percentage or something like that.

backwards compatibilty is there, also issue in #22258 is also addressed.
thoughts?

[Russell Keith-Magee]

If you're outputting to console, I don't see how you can display a progress indicator. Making a progress indicator dependent on the use of -o seems like a reasonable approach to me.

[Anubhav Joshi <anubhav9042@…>]

In f34e8fc890d44f4f913fe812f2cdeb5666f0fa27:

Fixed #22257 -- Added file output option to dumpdata command.

[Marc Tamlyn <marc.tamlyn@…>]

In 09ab447d083558c0822acd58ab621e06c206ccba:

Merge pull request #2465 from coder9042/ticket_22258

Fixed #22257 -- Added file output option to dumpdata command.

[Tim Graham <timograham@…>]

In e3c4205b79cff2be809c3698df2e9d53303b5070:

flake8 and doc fixes for refs #22257.

[ANUBHAV JOSHI]

Thanks for that.

[Gwildor Sok]

Thanks everyone for the quick implementation :)