[V6.0.2] No check when loading ir.exports & ir.exports.line

[Bernard Renier - ADN]

Hello,

In the context of our data migration work, we would like to standardize exports of data from OpenERP at user level. The aim is that user would be able to export object's data that he could reimport without any problem after correction (i.e. a customer list).

The easiest way to do that was to load tables ir.exports & ir.exports.line with predefined values. We did that but were facing issues with numerous values into ir.exports.line when loading a csv into these tables.

From a postgres standpoint all entries were created and were OK, but in OpenERP client (web and GTK) the predefined list does not appear (I mean when we loaded 15+ line items).

The issue is that the server does not do any checks on value that we load via csv into ir.models.line. If we had this kind of value check, it should be easier to determine why it has been loaded and why it does not appear on the client predefined exports.

Regards,

Bernard

[Vinay Rana (OpenERP) (vra-openerp)]

Hello Bernard,

Would you please elaborate more with proper example because I am not getting exactly.

Thanks.

[Bernard Renier - ADN (bre-adn)]

Hello,

I add here two csv used to load ir.reports and ir.reports.line. This is very useful for defining predefined csv exports on objects that you want to modifiy and reimport on a quite standardized way.

As you will see, the content seems OK and will be loaded without any error in DB. But when you go to the object list (res.partner in this case), it does not appear in the list of predefined exports.

This is due to the fact that system cannot interpret one or several values but no check/validation is done by OpenERP server at loading....

Regards,

Bernard

[Bernard Renier - ADN (bre-adn)]

Ir.exports line csv is here...

[Naresh(OpenERP) (nch-openerp)]

Hello Bernard Renier - ADN,

Thanks for reporting the issue. I had a look at both the csv's and they are absolutely correct i.e in OpenERP terms they are "not Import Compatible". So the server will easily import them without any problem. Now the case that you cannot see them when you are using the export wizard from both clients(web & GTK) . I did the same (note I am using Trunk version for both clients) from web I saw the predefined list. but not through GTK the reason is GTK loads the fields that are import compatible by default. (see the check box "import compatible") . Yes we need to improve here is when you uncheck this box then it should show you the predefined list.

So I would say we would improve this at GTK .

Correct me if I am missing something !

@community Team: can you please change the project and team to GTK ?

Thanks !

[Naresh(OpenERP) (nch-openerp)]

Hello Bernard Renier - ADN,

As I said above in my comment #4 the improvements are done @lp:~openerp-dev/openobject-client/trunk-bug-805243-nch. when you open the export wizard you will be only get to view the predefined list(If any) that are import compatible and if you toggle the import compatible check box then you will also be able to see the non import compatible list . i.e
checkbox == checked => Only import compatible list
checkbox == unchecked => All (compatible + Non compatible list)

Merge proposal link:https://code.launchpad.net/~openerp-dev/openobject-client/trunk-bug-805243-nch/+merge/67690

can you please provide your feedback !

Thanks,

[Bernard Renier - ADN (bre-adn)]

Hello Naresh,

I just saw our comments. Many thanks, this new approach on GTK is fine. I can't test it right now as I have no trunk environment at disposal here, but I would like to add that you missed one element of my original bug report:

"...no check/validation is done by OpenERP server at loading"

It means that you can put anything in ir.reports.line.csv and it is loaded by OpenERP without any error message.

As an example take the csv I sent, put "test" in one of the "name" columns, and you will see that the server will load the file into postgres without any error message, while this field does not exist in res.partner.

Regards,

Bernard

[Naresh(OpenERP) (nch-openerp)]

Hello Bernard Renier,

Thanks for your comments ! Yes I agree we need a check/validation at server when loading data in ir_export_line.
we will improve this at server soon !

Thanks,

[Bernard Renier - ADN (bre-adn)]

Many Thanks Naresh.

Have a nice day,

Bernard

[Olivier Dony (Odoo) (odo-openerp)]

Bernard,
The server part of this bug report was changed to wishlist, so this deserves a little explanation. Currently, all models in the ir.* namespace are meant for internal use by the system, and should not be explicitly accessed by users. Some are used by the administrator to setup access rights, but not more than that.
Integrators or experienced administrators are free to mess with these tables at their own risk, but these objects do not perform an extensive validation of their inputs, so this is considered a low-priority nice-to-have, not a bug.

Naresh,
Let's be very very careful, please triple-check your patch for the GTK client, this is the typical change that introduces regressions when trying to "improve" things that worked well enough in most cases.

Thanks!

[Naresh(OpenERP) (nch-openerp)]

Hello Olivier,

I had tested it well and still I would recheck it again for any possible regression.

Thanks

[Bernard Renier - ADN (bre-adn)]

Hello Olivier & Naresh,

I reported this bug because generating a standardized csv from OpenERP is very complex for a lambda user.

That's why I used ir.reports and discovered this issue (I tried also a lot of csv export modules on apps.openerp.com but no ones were on a completed stage or easy to use).

I completely understand this solution based on ir.reports is technical and has no priority...

Thanks to the transactionnal export from GTK/Web, exports can be done on csv, but this functionnality has two issues:
- You can't never reorder chosen fields (a nightmare when you have 15 fields and you forgot only one important field and have to put it on the end)
- You can select a same field twice because there is no indication about the fact that a field is already selected on right panel.

Ideally, the best csv export interface should be:

- You should see the list of fields on the left panel (as it exists currently) but you should have checkboxes that transfer fields to right panel.
- You should be able to reorder fields on the right panel

But all this is another improvement & useability story...

Regards,

Bernard

[Olivier Dony (Odoo) (odo-openerp)]

On 07/14/2011 06:34 PM, Bernard Renier - ADN wrote:
> Thanks to the transactionnal export from GTK/Web, exports can be done
> on csv, but this functionnality has two issues

You'll be happy to know we're preparing a new UI for CSV import in the
future, based on a preview of the first line of the CSV file :-)