A generic fifo and fifosock wrapper, under telnet and http server #2650
Conversation
There was a problem hiding this comment.
It sure took some time to get a catch on the fifo and fifosock logic.
But now I think I like it. Will use in other projects. Especially the fifo part.
Thanks for this great contribution.
As a general remark I prefer more speaking names so I see immediately what is meant and don't need to read the code first. (which is hard without clear naming again)
|
Thanks much for the feedback, @HHHartmann, and I'm glad to hear you think the modules are useful. :) |
|
Revised commits improve documentation a whole lot (by adding .md files, too) and now include the |
|
There's a conflict in The new |
|
OK, here's another stab. This one...
|
One occasionally wants a generic fifo, so here's a plausible implementation that's reasonably flexible in its usage. One possible consumer of this is a variant of TerryE's two-level fifo trick currently in the telnetd example. Factor that out to fifosock for more general use.
If we just pushed a little bit of data into a fifosock that had idled, wait a tick (1 ms) before transmitting. Hopefully, this means that we let the rest of the system push more data in before we send the first packet. But in a high-throughput situation, where we are streaming data without idling the fifo, there won't be any additional delay and we'll coalesce during operation as usual. The fifosocktest mocks up enough of tmr for this to run, but assumes an arbitrarily slow processor. ;)
Switch to fifosock for in-order sending and waiting for everything to be sent before closing. Fix header callback by moving the invocation of the handler higher
|
Some minor tweaks in response to out of band commentary. Notably, the |
Fixes to nodemcu#2650
Fixes to nodemcu#2650: - Convert fifosock to returning tables containing ctors - Improve docs - Add a missed :on("sent", nil) in the http server
Fixes to #2650: - Convert fifosock to returning tables containing ctors - Improve docs - Add a missed :on("sent", nil) in the http server
devbranch rather than formaster.docs/en/*.This pulls some things that I've had brewing in https://github.com/AcmeTensorToys/esp-lua_core for a while up to a more wide audience. It is mostly a refactoring of @TerryE's two-level fifo trick from the
telnetserver example into a more general module.While here, though, move the
httplua server example over tofifosockto guarantee in-order sending and waiting for all bytes sent to be queued; see #2647.This particular repackaging is almost entirely untested; there was some churn in liberating things from my particular environment. Feedback is more than welcome, and I will be doing some testing as time allows, but I wanted to solicit early commentary. :)