New import/export interface for 1.0

[kennethreitz]

• data.export('json') instead of data.json.
• data.import('json', value) instead of data.json = value.

[felciano]

Why the change? I thought the current version was very elegant.

[ghost]

This is a much needed change for two reasons: It's rather cumbersome to code an interface that lets you choose from a variety of different formats, because each module must currently be invoked as an object attribute. Also, this is needed to add support for import options, such as turning headers on and off.

[kennethreitz]

@teilo 👍

[iurisilvio]

I like this interface to support several formats, but you just can't use import as a method name.

[iurisilvio]

I tried a first draft of this issue in branch import_export because I really need custom configuration to import/export.

I don't like the import_ name, but used it to avoid naming discussion.

It is backward compatible and should work with all registered formats.

>>> s = "a,b,c\nd,e,f\n"
>>> d1 = tablib.Dataset()
>>> d1.csv = s
>>> d2 = tablib.Dataset()
>>> d2.import_("csv", s)
>>> d1.export("json") == d2.json
True

This way, it is easy to configure importers and exporters.

>>> d = tablib.Dataset()
>>> d.import_("csv", "a;b;c\nd;e;f\n", delimiter=";")
>>> d.dict
[OrderedDict([(u'a', u'd'), (u'b', u'e'), (u'c', u'f')])]

[iurisilvio]

Any thoughts here?

I really think this is the way to go with tablib API. It easily integrates third-party formats and allow configurable inputs/outputs without creating new formats.

I'm already running my code in my reports and it fixed a lot of "can't use tablib here because of custom formats". CSV delimiters, large Excel files, import/export to custom formats (e.g. import data from SQLAlchemy and export to Google Charts).

[iurisilvio]

I'm thinking about this issue, maybe we can use the json module names.

dset.loads('csv', 'some,data')
dset.dumps('json')

After that, we can also support file descriptors with dset.load and dset.dump.

[pstch]

Is there still work done on this at the moment ? I'd be glad to help, I realized I use tablib really often and that the project might need some help (seems like the main developer is quite busy atm).

I'm not really an experienced coder, but I have good understanding of Python, and I work with CSV/import/export libraries since a few months now so I may be able to do something.

[iurisilvio]

I don't know why I didn't saw it before, but tablib has import_set/import_book, which can be improved to accept a format parameter and work the way we want.

The dset.import_ is not necessary. All import_set format functions start with a dset.wipe(). If you want to load your data from an external source, you'll use anything from the dataset you created.

>>> tablib.import_set(text)  # auto detect format
>>> dset = tablib.import_set('{"my": "json"}', format='json')
>>> csv_text = dset.export(format='csv')

[iurisilvio]

The #191 PR was merged and I think this is done. Also,

Maybe it deserve better docs.

dataset.export('xlsx', freeze_panes=False)
dataset.load('a,b,c\nv1,v2,v3\n', format='csv')