Upgrade: v11.2.0

[kataras]

• Add context#FullRequestURI and NewConditionalHandler as requested at: Feature: Add Context#Request#FullRequestPath #1167 and Feature: Add applying rules on middleware #1170 by @zheeeng;
• Add warning messages when an end-dev missuses the MVC dependencies registration and controllers handling;
• Add an example to execute the ctx.Translate (using the i18n.Translate from /iris/middleware/i18n) inside the templates: {{call .tr "hi" "arguments..."}} when .tr is registered as a view data "tr": ctx.Translate, end-dev can use a view-data wrapper for common-used functions as they already do;
• Websocket: A lot of cleanup and performance improvements that can theoretically solve the C10M Problem
• Add tutorial for the Official MongoDB Go Driver.
• Add Context#ResetRequest and use it to update the request on core/handlerconv.FromStdWithNext as requested at How to add a value into the request context from a middleware function? #1180
• Do not call a controller's method if second output value of a dynamic dependency returned a not nil error as requested at Error handling in MVC mapper functions #1187
• Give the ability to modify the whole session's cookie on Start and Update/ShiftExpiration methods, add a StartWithPath helper as requested at About Session's Path #1186
• Add Context.ResponseWriter.IsHijacked to report whether the underline connection is hijacked
• Add the ability to intercept the default error handler by seting a custom hero.ErrorHandler to MVC application-level or per controller as requested at#1244, example at: https://github.com/kataras/iris/blob/v11.2.0/_examples/mvc/error-handler/main.go
• Extract the Delim for redis sessiondb as requested at sessions database redis keys delimeter #1256
• Fix hero/mvc when map, struct, slice return nil as null in JSON responses, as reported at Don't serialize MVC return nil #1273
• Fix pongo2 addons can't be registered, as reported at #1284 and view/django pongo2 conversions are now type aliases
• Full re-write of the websocket module;
  The v11.2.0 will contain a breaking change on the websockets usage, there is no other way to improve it. That one has its own repository for testing and experimentation to implement a faster and even more higher level websocket server and client written in Go which developers will be able select between gorilla/websocket and gobwas/ws for the underline protocol communication and eventually will live into Iris, click here to watch the progress of the new Iris websocket capabilities and features if you can't wait more.
• Replace all static methods with iris.Fileserver package-level function and Party and APIBuilder method HandleDir, read more at The new FileServer #1283
• The ctx.HandlerName() returns the current handler's function name of the chain. The new ctx.RouteName() returns the current route's name which is executing the handler. Also new ctx.HandlerFileName() which returns the exact program's source code position of the handler's function.
• Redis backend service changed from "redigo" to "radix" which is more active project and gives more options for connection pooling and clustering. The old /sessiondb/redis/service.Config was replaced by /sessiondb/redis.Config (no need to import other path for custom configuration now). Also a redis.Config.MaxActive allows to customize the max number of connections and the old redis.Config.IdleTimeout was replaced by redis.Config.Timeout which sets timeouts for redis reads, writes and dialing. Fix a bug which removed the whole session (like sess.Destroy() when using sess.Clear() with redis sessiondb.
• Implement a new View Engine for the jet template parser as requested at: How to use jet template in iris ？ #1281. Examples available at: _examples/view/template_jet_0 and _examples/view/template_jet_1_embedded
• Add mvc#Application.HandleWebsocket to register websocket controllers, its methods as websocket events now automatically, add a new stateless struct field tag as well.
• Add mvc#Before/AfterActivation.HandleMany and GetRoutes methods as requested at: How to use HandleMany in mvc #1292
• Implement a websocket stack exchange feature that allows scaling-out of websocket servers, add a built-in websocket.NewRedisStackExchange for scaling-out through redis pubsub system as requested at: [FEATURE REQUEST] adapters support for scalability neffos#3.
• Fix iris.WithoutBodyConsumptionOnUnmarshal not be respected on ctx.ReadForm and ctx.FormValues as reported at: ctx.Request().body only fetch once #1297
• New feature: Automatic Public domains for development using iris.WithTunneling configurator. Read more at: Automatic Public Domain for development #1305

Published

https://dev.to/kataras/iris-version-11-2-released-22bc

[jjhesk]

@kataras why sync map doesnt solve the problem?

[kataras]

You told me that it doesn't at our e-mail contact...I though it was buggy so I change it to our old-school way. Both ways should be safe, we already test it on running products and never got a datarace for about 2 years... I will need your code to test your case, please send me over the whole project and how you test it.

[speedwheel]

@kataras This example with the new mongodb driver came just in time.

I've been using Iris + MongoDb extensively over the past 1.5 years in my projects. But I never used the official golang driver since it was still in alpha. Now that it's in beta, I might make the switch for the next project, also I am curious to test its performance comapred with the community version driver and see if there is any noticeable difference.

[jjhesk]

@kataras here i just applied the new code server.go and i discovered more race errors than before.. please hold on this patch.

[speedwheel]

After doing some tests it seems like the community drivers are 2x faster than the official ones for selecting a single document.

10 vs. 20k req/second

[kataras]

@speedwheel That's awesome man!! I saw from its code that it may require more allocations to do things than the mgo driver but I myself, didn't expect so big difference between those two, should I add a tutorial for the mgo driver as well? It will be identical through, or should I leave it to you? 🥇

[kataras]

@jjhesk I know but I wanted to see if you get a difference data race stackrace, I replied to your new issue as well about byte buffer. Of course, the release is posponding until we fix all the opened bugs by the community, this includes the websocket datarace you mention, but still I can't re-produce that, can you make some code changes in your local iris and do further tests? It seems that you are deeply working on this for more than a week, you can open a PR or you could tell me what you learnt so far so I can continue investigating it. Thank you a lot, once again!

[speedwheel]

@kataras I opened an issue on jira about this to create more awareness regarding the performance side.

Yes I will do the tutorial, it might come in handy for some users, as for me, I will still continue using mgo for now.