03:23 <FromGitter> <Blacksmoke16> now on your way to `@bottom_level` :P

11:20 <FromGitter> <rjnienaber> more details from the reddit thread here (https://www.reddit.com/r/crystal_programming/comments/zx9yrs/create_a_passwordencrypted_db_accessible_in_a_lan/)

11:36 <FromGitter> <rjnienaber> I have a few questions about my project, amazonite (https://github.com/rjnienaber/amazonite), an unofficial AWS SDK. For AWS SDKs in some other languages (e.g. javascript, ruby), there is a monorepo but the maintainers publish separate packages for each API, e.g. aws-sdk-s3, aws-sdk-dynamodb, etc to the package manager's website (e.g. npm, rubygems) . As I understand it, Crystal dependencies are managed via

11:36 <FromGitter> ... the VCS (Github for the most part) so you download the whole repo when you add a dependency. ⏎ ⏎ So my questions are: ⏎ ⏎ 1) Is there a downside to including all APIs in the repo from a compiled code size point of view? Will the compiler leave out the parts that are unused to create smaller binaries? ... [https://gitter.im/crystal-lang/crystal?at=63aecd37c4fec572d65f508f]

11:44 <FromGitter> <oprypin:matrix.org> @rjnienaber: all user-facing code should be under src. can be src/generated. codegen looks good to be outside of src

11:46 <FromGitter> <oprypin:matrix.org> you can test any theory regarding whether compiler leaves stuff out. both in terms of compile times and binary size. try adding and removing dirs and recompiling. try using most classes and then not using almost any classes and compare recompiling.

13:04 <FromGitter> <moe:busyloop.net> oh i misunderstood, sorry. ⏎ but as the people on reddit wrote, it really doesn't sound like something you want to build yourself. would suggest to rather use one of the existing IT asset management tools (e.g. lansweeper or GLPI). or depending on your size/preferences, a plain old google sheet or airtable can go a long way.

13:11 <FromGitter> <moe:busyloop.net> much appreciated! an AWS SDK is really sorely missing from crystal. ⏎ (doing it all by hand with aws-signer.cr gets old quick...) ⏎ ⏎ godspeed to you. hope amazon will sling some sponsor-dollars your way when they see you doing their work. [https://gitter.im/crystal-lang/crystal?at=63aee39c3b37dd2c11562443]

13:14 <FromGitter> <moe:busyloop.net> > Is there a downside to including all APIs in the repo from a compiled code size point of view? Will the compiler leave out the parts that are unused to create smaller binaries? ⏎ ⏎ yes normally it does. i'd def start with just including all APIs - if it causes any issues later on then that can still be changed fairly easily.

14:56 <FromGitter> <rjnienaber> leaving `codegen` out of `src` seems like a good idea. i'm going to give that a go. thanks :)

14:57 <FromGitter> <rjnienaber> okay cool. thanks for the response :)

15:04 <FromGitter> <Blacksmoke16> if you structure the code so there's a file under `src` for each API. e.g. `s3.cr`, `dynamodb.cr`, etc. then the user could manually only require what they want

15:05 <FromGitter> <Blacksmoke16> e.g. `require "amazonite"` would be a file that requires each separate API file. but the user could also do like `require "amazonite/s3"` to explicitly require *just* s3 stuff

17:30 <FromGitter> <oprypin:matrix.org> in fact, consider not even providing a file that requires everything

18:35 <FromGitter> <michael-swan> I seem to be observing oddness relating to tuple literal parameter parsing: ⏎ ⏎ ```code paste, see link``` ⏎ ⏎ meanwhile adding a first argument: `f :foo, {1,2,3}` doesn't produce this error or adding parenthesis: `f({1,2,3})` also doesn't produce this error. [https://gitter.im/crystal-lang/crystal?at=63af2f7efb195421bd5fb425]

18:36 <FromGitter> <michael-swan> Am I doing something naive here, or is this a known/potential bug?

18:36 <FromGitter> <Blacksmoke16> basically same issue as https://github.com/crystal-lang/crystal/issues/7923

18:36 <FromGitter> <michael-swan> Huh interesting

18:38 <FromGitter> <Blacksmoke16> idk what the context is, but you could ofc do like `def f(*args : Int32)` then could just do `f 1, 2, 3`

18:39 <FromGitter> <Blacksmoke16> where `args` parameter would then be a tuple of the args you passed to the function

18:39 <FromGitter> <michael-swan> Gotcha, yeah no worries. I can find another way it's not a problem. Parsing a language such as this is pretty difficult I imagine

18:40 <FromGitter> <michael-swan> If it follows a comma it seems to not be ambiguous

18:41 <FromGitter> <michael-swan> I wonder how crazy changes to the parser would have to be to support differentiating cases such as this

18:41 <FromGitter> <Blacksmoke16> was mentioned in that issue, basically would need to support 3 level deep lookahead which is kinda :/

18:42 <FromGitter> <michael-swan> Yeah, interesting

20:15 <FromGitter> <benash> It seems to me that Crystal supports multiple dispatch. (I was excited to discover this code did what I wanted it to :) ): ⏎ ⏎ ```code paste, see link``` ⏎ ⏎ But I didn't see this awesome feature mentioned much in the documentation. Are there any limitations to be aware of? The Julia community talks about its own support being hugely helpful for code reuse. [htt

20:15 <FromGitter> ... ps://gitter.im/crystal-lang/crystal?at=63af46f951d2482e91e4a0cc]

20:16 <FromGitter> <Blacksmoke16> basically its just https://crystal-lang.org/reference/1.6/syntax_and_semantics/overloading.html

20:26 <FromGitter> <benash> OK so the docs touch on it there in the section on overloading—but it's overloading that dispatches based on an arbitrary number of runtime types, right? It seems that very few languages do that.

20:35 <FromGitter> <Blacksmoke16> thats my understanding

20:49 <FromGitter> <benash> Pretty cool