<asterite>
In that case, you can merge the ssl -> tls PR. Could you add that as a breaking change to the changelog, plus all the other changes that were made to openssl? I think there are a few more breaking changes, the split between openssl socket, I think
<jhass>
yes, have that locally stashed away already
<asterite>
\o/
<crystal-gh>
[crystal] jhass closed pull request #2797: Replace and rename all mentions of SSL with TLS (master...ssl_to_tls) https://git.io/voZac
<asterite>
Great! After that, I think I'll prepare the 0.18.0 release. This time I have a bunch of medium projects I compile and run specs to verify that none broke, for example crystal-pg, shards, kemal, crystal-sqlite3 and a few others... they all seem to compile fine, so I'm hoping we won't need 0.18.1 and others... but who knows :-)
pawnbox has quit [Remote host closed the connection]
<jhass>
:)
pawnbox has joined #crystal-lang
<jhass>
asterite: btw british or american english in the changelog and docs? :P
pawnbox has quit [Remote host closed the connection]
pawnbox has joined #crystal-lang
<crystal-gh>
[crystal] jhass pushed 1 new commit to master: https://git.io/vol89
<crystal-gh>
crystal/master fdbf9f3 Jonne Haß: update changelog for TLS related changes
<asterite>
I always mix them because I don't know the rules
<asterite>
I use `z` in initialize and others
<asterite>
but I write behaviour... that's british?
<BlaXpirit>
>> abstract class Base; @a : Int32; end
<asterite>
i'll try to write behavior from now on :)
<DeBot>
BlaXpirit: instance variable '@a' of Base was not initialized in all of the 'initialize' methods, rendering it nilable - https://carc.in/#/r/11j4
<BlaXpirit>
abstract classes shouldn't even have initialize
<jhass>
well, they may provide a default one to their childs
<asterite>
True, but since `@a` will need to be initialized, I prefer it to be initialized there
<asterite>
I think we discussed that in the past, but I don't know if in an issue...
<asterite>
basically, define initialize(@a) in Base, and use super(a) in Sub
<asterite>
but we can leave it open for discussion
<BlaXpirit>
hm I seem to remember having some problems with that. gotta try again
<crystal-gh>
[crystal] asterite pushed 1 new commit to master: https://git.io/vol4r
<crystal-gh>
crystal/master 059c462 Ary Borenszweig: Missing tiny rename from `ssl` to `tls`
<asterite>
it's used in the other auth token, mac, but in bearer it's ignored
<asterite>
Hmmm... I'd like to try merging #2745
<asterite>
I didn't try ysbaddaden's solution, but I'm almost certain it should crash because of threads + gc
<asterite>
plain old getaddrinfo should always work, even if a bit slower
<jhass>
fine with me, can't make the situation worse than now and it's nothing that requires breaking API to improve
ome has joined #crystal-lang
<BlaXpirit>
yeah, fighting with this "was not initialized" thing is by far the most annoying. sometimes I just wish I could tell it to shut up and just leave it uninitialized
<asterite>
:)
<asterite>
That's the easiest thing to do from a compiler's perspective too
<asterite>
just have everything be nilable, let the use handle it all the time
<BlaXpirit>
>> class C; a = uninitialized Int32; end # wishlist
<BlaXpirit>
also, I said "slower" above - well I meant "faster" - the faster job should have 1 as the result. what confused us: the name should be .slower_by, not .slower
Oliphaunte has quit [Remote host closed the connection]
fryguy9 has quit [Quit: Leaving.]
Dreamer3 has joined #crystal-lang
Dreamer3_ has joined #crystal-lang
Dreamer3_ has quit [Max SendQ exceeded]
Dreamer3_ has joined #crystal-lang
fryguy9 has joined #crystal-lang
<paulcsmith_>
Is what I'm trying to do here possible with Crystal? https://gist.github.com/paulcsmith/9771a0a57eb400781e43c9da583aae00#file-router-rb-L15 I'm trying to add some routes and their handlers. If there is a match then it instantiates a controller and I attempt to call a method on the controller dynamically. I'm not sure that's possible though
Oliphaunte has joined #crystal-lang
<paulcsmith_>
So it's clear `send` doesn't exist, but is there some other way to do what I'm attempting to do?
<BlaXpirit>
paulcsmith_, at the lowest level, this can't be dynamic, so you basically have to write a case statement and call methods depending on a condition
<BlaXpirit>
the way it's mitigated is with macros and other compile time constructs
<BlaXpirit>
paulcsmith_, for example, you could list all methods of a class and generate a branch for each of them, or something of that sort
<paulcsmith_>
BlaXpirit That makes sense. Thank you :)
<BlaXpirit>
paulcsmith_, jumping into implementing this pattern is difficult for a newcomer. you can check out existing projects, like Kemal
<paulcsmith_>
Good idea. I'll check out Kemal's implementation
bcardiff has quit [Quit: bcardiff]
<crystal-gh>
[crystal] will opened pull request #2833: Fix benchmark spec (master...benchmark-spec) https://git.io/vo853
<paulcsmith_>
Yeah I'll check that out, but I maintain a few Elixir projects that are keeping my busy. Maybe I can find some time on the weekend though :)
<BlaXpirit>
jhass, that's nothing, check this one https://crystal-doc-new-design.herokuapp.com/Exception.html#new%28message%3D%3Cspanclass%3D%22n%22%3Enil%3C%2Fspan%3E%3AString%7CNil%2Ccause%3D%3Cspanclass%3D%22n%22%3Enil%3C%2Fspan%3E%3AException%7CNil%29-class-method
daneb has quit [Remote host closed the connection]
<BlaXpirit>
and it's my fault even, i used html as anchor
<jhass>
hehe
<BlaXpirit>
but previously there were conflicting anchors or something like that so it was impossible to link
<jhass>
but yeah, we probably should just sha1 that and then even the first 8 characters should be good enough
<jhass>
good enough for git, good enough for us
<jhass>
extremely unlikely to collide on a single page
<jhass>
we can still keep the "verbose" anchor available for stuff that want's to generate links or so
<BlaXpirit>
jhass, I think it would be best if the link actually told anything about what you're going to. just a hash is unacceptable
<BlaXpirit>
paulcsmith_, well they have overloading on just the number of arguments, that's good for them but ..
<paulcsmith_>
Oh right Good point. Well maybe in the URL it gives overloaded functions a hashed id. So it would be something like /api/Crystal/Macros/TypeNode.html#method_name-hashedid That way it's shorted, still shows what it's for and works with overloaded methods. Maybe that'd work
<paulcsmith_>
The URL wouldn't include what the overloads are, but that wasn't possible before without decoding the URI
<crystal-gh>
[crystal] jhass closed pull request #2820: Update cross compile docs, flags docs and use flag? instead of ifdef in the examples (gh-pages...cross_compile_flags_ifdef) https://git.io/vonpX
bjz has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
matp has joined #crystal-lang
Oliphaunte has quit [Remote host closed the connection]
Raimondi has quit [Quit: All hail WeeChat 1.5-dev!]
matp has quit [Ping timeout: 264 seconds]
<crystal-gh>
[crystal] jhass pushed 2 new commits to master: https://git.io/vo4Y2
<crystal-gh>
crystal/master 2705907 Celso Fernandes: Update to use compile instead of build command...
<crystal-gh>
crystal/master e983ab1 Jonne Haß: Merge pull request #2760 from fernandes/use_compile_on_travis...