Hi, im trying to add some exception monitoring, since castra is catching all exceptions and does its own stuff to it i can not just use the JVM setDefaultUncaughtExceptionHandler You think its ok if wrap-castra has an optional exception-handler option which has to take the exception as arg, and return the exception back (after you are done with your side effects)

i have on the frontend on-error with mkremote but at that time im already on the frontend and have to callback to the backend and also the exception is not in its original form anymore

actually it would be cool to also get fn-name and args. maybe an on-error on middleware and have something like

(catch Throwable e
           (when on-error (on-error fn-name args e))
           (response body-keys req {:error (ex->clj e)})

@peterromfeld: i just pushed off the quickfix i was using last night. https://github.com/jumblerg/castra/blob/master/src/castra/middleware.clj#L126

basically, i'm printing off the handled exceptions to stderr before they're returned back to the client.

this should cause them to appear in logs, etc. but i'm not certain this is the optimal approach.

any reason why hoplon doesn’t use the history functionality from google closure for route cells?

i've not looked at google closure's api, but does it wrap the html 5 pushstate api?

probs

a better question might be what would we gain from using it?

i’m not seeing my route cell updating in response to setting the window hash in phantomjs

but the bidi router did (based on goog)

do you happen to know, offhand, if the route cell works in other webkit implementations?

works fine in chrome

just tinkering now

i got a test that looks like

i believe webkit and chrome parted ways, chrome uses the blink fork now i think. but the route cell was implemented pre-blink probably.

(deftest ??handler?=
 (let [h?= (handler?= :home)]
  (navigate! :home)
  (is @h?=)
  (navigate! :privacy)
  (is (not @h?=))))

passed with a bidi router but not with a route cell

i’ll give safari a shot

yeah, works in safari

i got a (j/cell= (prn route=))

i see the printout in chrome and safari

don’t even see it in tests

navigate! is just a wrapper around bidi and

(defn set-hash!
 [s]
 {:pre [(string? s)]}
 (-> js/window
     .-location
     .-hash
     (set! s)))

actually @jumblerg it looks like there is a Html5 and regular Html version of the History management from google

https://github.com/google/closure-library/blob/master/closure/goog/history/history.js

looks like it has tons of browser support code to make navigation work

will use the Html5 pushstate api if available

makes sense to adopt that into Hoplon or make a separate route-cell lib?

btw @jumblerg printing stack trace by default is not a good option 🙂 Most serious deployments use a logger of some sort. You might also want to increment all kinds of stats so best to expose that via something like an on-error callback

@dm3 well i think i have to do something to get it working, i’ll let you know how i go 🙂

@dm3: wouldn't exceptions written to *err* be consequently handled by any logging framework?

you mean like system-level logging framework

this is not my area of expertise; i started into a deep dive on this yesterday, but cut myself short since i had to ship some code. this was a stopgap solution.

like syslog

I mean JVM ones like logback or log4j

yeah, that's what i'm using.

no, they don't pick up anything you don't give to them directly

e.g. via clojure.tools.logging

i think there's a with-logs fn or something similar than will capture it, but i agree that @peterromfeld's approach is probably a better solution.

you should be able to inject a middleware in the RPC chain

that'd be a more general solution

i suspect there's already a technique for doing this that i'm just not aware of; i find it hard to believe that this case isn't already covered somehow.

a middleware that grabs the castra-handled error off the response and logs it?

there's probably something like this out there already.

yeah, I haven't done much with castra

but it seemed like Castra just serialized all of the caught exceptions as a response and never gave a chance to do anything else with them, no?

yeah, the exception gets caught and passed up to the client wrapped in an http 200, where it is inflated back into an error.

this is handy in many cases, for example, authorization errors (404s), where you want the client to be able to handle the error by reverting to a login screen.

You can make a Middleware but still someone might want the original exception before castra dfl it

i would love to see a on-error option on castra middleware, which calls fn-name/symbol + args + exception on this function 😄

@peterromfeld: it seems that something like this would be useful; will definitely make sure we get in right before merging into master. pull my branch and make a pr off of it if you like.

ok ill look into it tomorrow if i can come up with some elegant solution

i want to circle back with micha or alan and make sure i'm not missing anything first; there's likely some less obvious technique for doing this already i'm unaware of.

maybe some implicit and undocumented change to the environment that may cause certain classes of errors to propagate up into logs instead of being sent across the wire to the client when in production.

you can have a defaulthandler for uncaught exceptions.. but castra catches it 😉

that's not a great solution either - you should be able to access the inputs that caused the exception too

yup i also would like that too, more context better for debugging later

@dm3 this seems to work

(defn history-cell
 []
 (let [c (j/cell nil)
       history (History.)]
  (j/with-let [_ (j/cell= c #(.setToken history %))]
   (goog.events/listen history goog.History/EventType.NAVIGATE
    (fn [e]
     (reset! c (.-token e))))
   (.setEnabled history true))))

it works a bit different to a route cell though

you reset it to navigate

and yep, that totally fixed my tests

@jumblerg @dm3 there must be some magic x-browser something in google history that route cell doesn’t have

because my phantomjs tests are passing now

Has anyone used the w2ui library with joplin?

hoplon

@jumblerg sorry been under the gun here at work today, firefighting

i'm sure... no worries and no rush at all.

had a partial redis meltdown

you finally broke it!

no radioactive wazste was leaked

congratulations?

haha yse whew

yes

redis is the best software ever

antirez deserves free beers wherever he goes

well, you've been trying to break it for a long time.

in perpetuity

man

we should like, send antirez a christmas bonus

seriously

i guess you deserve a few yourself in light of your most recent success.

i bet james is down

lol

yes, we're doing it

@chopmo: welcome!

Thanks @flyboarder !