Description

mohsen1

opened on Feb 22, 2017

Issue body actions

I did try putting together a PR that runs all tests at github/fetch but got overwhelmed by amount of work required. Basically it should be done via Git submodules and a shell/node script that swaps the implementation in the submodule with dist of this project.

Doing most of things manually I saw a few tests failing.

Reacted by Patrick Burtchaell, Steven and Álvaro Torralba

Activity

1. developit commented on Feb 22, 2017

   

   developit

   on Feb 22, 2017

   Owner

   More actions

   Sound good. We will definitely need to selectively ignore some tests though.
2. mohsen1 commented on Feb 22, 2017

   

   mohsen1

   on Feb 22, 2017

   Author

   More actions

   Which parts you think it should fail?
3. developit commented on Feb 22, 2017

   

   developit

   on Feb 22, 2017

   Owner

   More actions

   unfetch intentionally doesn't recreate the Headers or Response classes, instead using POJO's for them to shave bytes (so those tests should all fail). The rest are just size tradeoffs - I was on the fence about even supporting Response#xml() and Response#blob(), but the size bump was minimal (10b) for them. To me, unfetch is best kept as a bare minimum feature set: loosely emulating the fetch() API semantics but allowing the same userland usage so that it remains transparent for the 99% case. Anything that falls under the umbrella seems like fair game.

   Reacted by Mohsen Azimi, Patrick Burtchaell, Paul Martin, su, Tung Nguyen, Steven, Rajesh Segu and Álvaro Torralba

Issue actions