Possible inconsistent method: parse

I’ve run into some trouble while trying to customize the way response bodies are parsed – specifically on the client.

From these tests, it appears that the parse method of the node library is used to customize how response bodies are parsed.

This test shows that the parse method of the client library is used to customize how request bodies are serialized.

Is this inconsistency intentional? If not and a maintainer would like to specify the intended behavior, I’d be happy to open a PR.

[kornelski]

The term serialize is used in Superagent’s codebase, so for 2.0 I suggest:

• parse(fn) for parsing response body only (this is a breaking change)
• serialize(fn) for generating data to send in the request

[focusaurus]

Sadly, it looks like request.parse indeed sets a request body serializer and then this._parse in the browser request instance is indeed a function used for serializing request bodies. Thus I don’t think the confusion is request vs. response but rather parse vs. serialize. I’d like to see a PR for the following changes:

• rename the private var this._parser to something more clear like this._serializeBody in the browser client code
• Add a new clear public API to set the request body serializer. Perhaps req.serializeBody() (open to more concise suggestions for the name). This can be an alias of the current req.parse in the client code.
• I don’t see any functionality in the node code to provide a custom body serializer. It seems to all just be hard coded in exports.serialize which presumably can be overridden in a straightforward way if a user has read the source, but maybe we should make an explicit API to keep parity with the browser code.
• deprecate existing request.parse with corresponding doc updates