@theindigamer i rewrote the introduction, updated the tutorial, added an execution model section, and properly wrote the guide to rest of lwt in the new manual draft. rendered it to html, take a look here https://ocsigen.github.io/lwt/manual-draft/Lwt.html
would be interested to see if this is helpful, and in any other feedback you may have :)
I finished reading the sections before Sec. Rejection. Comments:
- Links to lwt functions are dead.
- In Sec. Execution Model, the word "resolution" is used as a noun but it isn't mentioned before. Do you intend to use I/O promise instead?
In Sec. Waiting, should
If p1 is fulfilled now or later with value v, f v is called. f v might be called immediately, or after some time.
instead have something like "
f vwill be called when p1 is fulfilled"?- "Actual promises are read-only.", it's not fully clear what that means ... it can't mean that promises are immutable (because you can resolve them through the resolver). Are they read-only modulo resolvers (which are typically hidden away and not used directly)? It might be better to have a few more words here :smile:.
Apart from these, I think the rest of it is a good read.
Also, for some reason, the deprecated section is not linked to in the sidebar.
@theindigamer: thanks. the links to anything outside this module are dead, because it's just one file uploaded for now. i suppose i should put the rest of the current manual in the same directory. also, ive since rewritten most of the function docs. i'll incorporate your feedback :) i hid the deprecated section on purpose, but ive since also put it back, without linking its subsections in the sidebar
basically, as suggested, only the intro parts of that were ready for consumption :)
Orbifx Anton Bachin (Gitter): is this how Lwt implements Lwt exceptions? https://hal.inria.fr/inria-00432575/document
(and separate the monads)
this basic error monad carries around
exn for errors, an open variant type
there is also
Lwt_result, which provides a separate error monad
with it, the user can choose what types to use for carrying around errors. they could be
string, or some inferred polymorphic variant, or an open variant, or anything else
yeah, just being thorough :)
see discourse topic here https://discuss.ocaml.org/t/lwt-3-2-0-released-we-plan-to-change-lwt-bind/1337
Actually, I've learned them both.
Lwt_log is easier to start with; Logs is better to elaborate on )
Orbifx: i don't know, it's the default room created by gitter
two of the plots in this graph effectively compare performance of lwt 3.1.0 to 3.2.0 https://github.com/ocsigen/lwt/pull/500#issuecomment-343535322
it looks unchanged
in my own measurements while making changes, the effects were something like small-percentage slowdowns of bind. given that bind is by far not the most common or time-intensive operation performed in an lwt program, i also expected there to be no real effect
@argent-smith i believe all the breaking changes are the ones listed in the changelog here https://github.com/ocsigen/lwt/releases/tag/3.0.0, also you can see details here ocsigen/lwt#308
if anything else broke, it's an accident, and we are not aware of any such accidents
Orbifx Ocsigen*
smondet both are low traffic, but stuff still happens
Orbifx Anton Bachin (Gitter): is there a particular reason for choosing MIT? something tactical?