Columns [meta] and [type] included in mssql query

[rnam415]

When connecting to a SQL Server instance, [type] and [meta] columns are ending up in the raw sql query.

"errors": [
{
"status": "500",
"code": "EUNKNOWN",
"title": "An unknown error has occured",
"detail": "Something broke when connecting to the database - Invalid column name 'meta'."
}
]

SELECT [id], [type], [meta], [first_name], [last_name], [phone_number], [middle_name], [title], [suffix], [email_address], [birth_date], [date_created], [date_modified], [address_line1], [address_line2], [address_line3], [address_line4], [city], [state], [postal_code], [country], [clark_region] FROM [workers] AS [workers];

[paparomeo]

We have never tested with MS SQL Server. It looks like meta is a reserved name in SQL Server and we'll need to escape it. I'll try to have a look at this, but meanwhile if you want to submit a PR with a fix it will be welcome.

[mrrusof]

The same happens when you use postgres. Is inclusion of attributes type and meta intentional? If so, what is the purpose for each one?

[mrrusof]

Cool project btw!

[rnam415]

I did not pursue this issue any further

[pmcnr-hx]

MSSQL is now available on Linux so I'll try to reproduce this soon and hopefully add MSSQL as unofficially supported (not sure if it is possible to run it on Travis).

[aledalgrande]

What about postgres? Is it possible to exclude these fields when using that dialect? I don't know why it's trying to pass a value to a type column, which is not needed.

[pmcnr-hx]

I need a better look at the code to refresh my memory. I'll try to get round to it today.

[aledalgrande]

Trying to look at this one too, it is a blocker for me right now.

[championswimmer]

@aledalgrande what exactly is the problem in postgres ?

[aledalgrande]

@championswimmer when I call a resource creation endpoint I see this from postgres:

ERROR:  column "type" of relation "<my_resource_name>" does not exist at character 34
STATEMENT:  INSERT INTO "<my_resource_name>" ("id","type") VALUES ('c3a32f92-bcbf-4f48-8a34-9747417d628c','<my_resource_name>') RETURNING *;

I don't have the type column, so it fails. Why is it trying to save it? The type shouldn't need to be stored right?

(keep in mind I create my own models with sequelize separately)

[championswimmer]

@aledalgrande done

[paparomeo]

I'll have a better look at this during the weekend and see if it fits with the design philosophy of the jsonapi-server handlers. However, the idea for the data store adapters / handlers is that the model is generated by the handler and should be opaque. When you already have your own model generated externally the recommend approach is to write a custom handler.

[fny]

So I'm having this same issue using sqlite, it's querying for meta despite me never specifying it as an attribute. Is there an expectation that the table contains both a type and meta column?

[paparomeo]

The short answer is: yes, as explained in my comment above, The structure of the table is automatically generated by the handler and should be opaque to applications. Also, SQLite is not officially supported at the moment. 4 of the automated tests are failing for SQLite and until those pass we cannot support it officially. Contributions to help figure out what the problem is and fixing it are welcome.