Extend response_fields to support field exclusion

[merkys]

Currently response_fields query parameter allows the user to explicitly enumerate fields to be included in the response. I suggest a mechanism to explicitly exclude particular field(s) using the same response_fields query parameter.

Now response_fields=a,b,c means that the user requests response with only a, b and c fields. As field names cannot have - sign, response_fields=-a could be defined to stand for all fields excluding a, and response_fields=-a,b,c - all fields excluding a, b and c etc.

An alternative proposition is to introduce response_fields_exclude query parameter. However, such introduction would also require description of how response_fields and response_fields_exclude interact.

[rartino]

Can you give a scenario where this would be useful?

My thought process is that this feature would just cater to bad habits. Sure, we all like to write less (give me -c instead of a,b,d,e). But, if you are trying to save bandwidth (and what other reasons would you have to use response_fields?) and you start saying "everything except c" you are going to get, e.g., database-specific fields, or fields from newer version of the standard that you are anyway going to throw away.

Regarding response_fields_exclude, well, also in your proposal you need to describe how multiple (+) and - entries interact. Is it an error for the same property to appear more than once? Or does the last (or first) take precedence?

[merkys]

Can you give a scenario where this would be useful?

My thought process is that this feature would just cater to bad habits. Sure, we all like to write less (give me -c instead of a,b,d,e). But, if you are trying to save bandwidth (and what other reasons would you have to use response_fields?) and you start saying "everything except c" you are going to get, e.g., database-specific fields, or fields from newer version of the standard that you are anyway going to throw away.

The reason I came up with this proposal is usually huge amounts of data returned in cartesian_site_positions. In cases when I am sure I won't need the coordinates, it is convenient to ask for everything except that.

Regarding response_fields_exclude, well, also in your proposal you need to describe how multiple (+) and - entries interact. Is it an error for the same property to appear more than once? Or does the last (or first) take precedence?

The alternative proposition of response_fields_exclude should serve only as a motivation for my primary proposal. That is, it just showcases the problems arising when using yet another query parameter instead of response_fields.

[merkys]

This issue should be offset by solving #184