Player is loading...
Scala best practices I wish someone'd told me about
Nicolas Rinaudo, CTO of Besedo
Thursday, June 13, 2019 · 3:47 p.m. · 48m 48s
Scala is a language with many qualities, but it has more than its share of quirks and small things that don't exactly behave as you'd expect. This talk is about the more commons gotchas in the language, and the standard recipes we have for dealing with them. A lot of them are assumed to be common knowledge and never really brought up, which means that a lot of Scala developers had to find out about them the hard way. The purpose of this talk is to try and save some pain by puting the worst offenders in the spotlight and, hopefully, starting conversations around them. Required Knowledge: Some familiarity with the language, while not required, would help get the most out of this talk. The best practices are explained in terms of the traps they allow us to avoid, and it can be hard to understand these traps without at least some basic understanding of the language.
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:00
alright so um had this afternoon to talk to you about this fact is it's gotta
00:00:05
and just to give you a bit of background on me so that you understand why
00:00:08
i actually can't talk about these things and is it your company called this it though
00:00:13
we recently murdered free completely different scenes one doing node
00:00:17
wondering to um java and wondering scholar and we're trying to
00:00:21
come together with the same processes and technologies it's working out about as bad as you'd expect
00:00:28
but we have finally agreed that's kind that was probably the language we all wanted to use
00:00:33
and this is getting a pretty interesting perspective because i'm getting to see talented developers that
00:00:38
nothing about scallop or function programming struggle with
00:00:41
learning the language and i'm rediscovering all the things
00:00:45
is that really trouble me as i was running the language so i'm saying lots of problems
00:00:50
that keep coming up and i'm writing down these problems and solutions to them and this presentation
00:00:56
is um the best ones the most common once the most interesting ones
00:01:01
or is it completely in knots one but i want to chat was it
00:01:05
so the way it's going get well okay is i'm going to show you some code
00:01:10
you are going to have to work out what i think is wrong with the code
00:01:16
explain it to me if we can refer don't have the time or just give the solution i'm going to tell
00:01:20
you what's wrong with it how to work around it and
00:01:23
uh in the remote next one so um let's get started
00:01:31
it's a very first bit of code that i want to show you is this so how how this work again right so what
00:01:38
why do you guys feel that i don't like this code i'm just comparing to raise is gonna be the hand um
00:01:46
often the to all the slides and is a good idea about what
00:01:50
the answer is going to be so yeah it quite obviously the problem is
00:01:54
with a re compare reason can anyone tell me why this is probably not great
00:02:02
okay
00:02:03
okay uh the main problem that is that it doesn't actually work
00:02:08
the rent is sent only collection invent highest colour
00:02:12
collection library start doesn't have a working equals method
00:02:15
so comparing always like that is going to not work ever and
00:02:21
we know that's when it's kind of approach we have had problem before
00:02:24
so we don't use a razor when we do we don't contend that way
00:02:28
well to begin is to people who come from a java background or javascript background or any other background
00:02:34
this doesn't make any sort of sense this is completely nuts um the reason doesn't work you have to do with the
00:02:39
way to go to the j. v. m. arrayed have reference
00:02:42
equality not value quality it's not something interesting just doesn't work
00:02:47
so instead of using equals fall arrays use same elements and again you have to
00:02:54
know that and it's going to but you sometime if you don't know it so
00:02:58
actually to yeah and if you didn't know about that great adjusted you i was of painting production
00:03:05
right so that was my first problem with every compare
00:03:10
regions and given the fact that this is kind of tennis
00:03:14
and also going to try and for anything one of this problem explained how
00:03:18
seine are fixed or improved or not at all in later version of colour
00:03:24
and so since four races can actually that's thirteen has immutable raise which behave
00:03:30
saintly with because we call them not some stuff so if using it will raise perfectly fine
00:03:35
you too will raise even dotty thought i can tell a completely nuts and and try to stay away from them if you can
00:03:45
right so this bit of code which looks perfectly fine dining many have any clue what i don't like it
00:03:53
okay
00:03:56
right so this is not the first and somebody giving the answer probably should changes like this
00:04:00
is completely correct but not the point to the slide um and again there's often a hint
00:04:07
um the front hey it's start i haven't type annotated my method so you could return anytime
00:04:13
in the components construct infrared and the best time again for the best type you can infer here
00:04:19
what you'd expect it to be option ave especially given the fact that i haven't a my method as option
00:04:25
it's actually going to be some which is not a massive they'll most of
00:04:31
the time if you have a summit to subtype of options so it's alright well
00:04:35
it could be a problem say if this is a library and i were to probably
00:04:38
starts then the type of the method actually changed and you break binary compatibility just like that
00:04:44
so when you have public methods public members
00:04:49
type annotate them because otherwise you will eventually break something and the types are inferred
00:04:55
are not actually the type you think or inferred watches not was not all the time
00:05:00
so just fix it that way put an extra seat type annotation and then everything works out just fine
00:05:07
this is kind of improved is kind of really 'cause implicit members have to have a type annotation
00:05:13
but at the same time into this it's are being deprecated things kind of result does actually change all that much
00:05:18
so this is still pretty much of a problem in to that city names got free as far as i know
00:05:24
and due to type annotation on any of your public members
00:05:33
you get operators
00:05:36
this is problematic cable ten why so first what is good supposed to do um this syntax here
00:05:44
is expected to read created topple who's left member is one and who's right members for divided by to
00:05:51
and this is while what happen if you read that using the ascii operators
00:05:57
and the documentation tells you that you ask operators and there you could equivalents
00:06:01
are exactly the same pair as they are in fact not the same says does not come how well if
00:06:08
you want to use the ascii operator it complies just fine so it was a problem here is very subtle
00:06:14
the problem has to do with operator priority is kinda where the rules the rules are the
00:06:20
way than hardcoded but if you operator starts with an eye for an eye with a minus
00:06:25
so and it's going to have um lower priorities and say divide
00:06:30
but this correctly here does not start with a minus
00:06:34
so is not actually the same operator as you ask hero
00:06:38
contrary to what the condition expressly states so this is very confusing and it
00:06:43
is but uh most of the time to use yes q. operator instead even though
00:06:48
it looks much nicer within a good characters um
00:06:53
there are ways to still have a nice display you can use fonts with
00:06:56
teachers you can use some advanced features such as he might have pretty printing
00:07:01
uh but it's not going to give him a problem much longer anyway because it's deprecated into that thirteen so
00:07:07
if you guys like me had rules to turn all you ask your bridges into you could
00:07:11
you could you like that you're going to have to revert that 'cause it's not going to find
00:07:16
uh_huh
00:07:20
any variations what is wrong with the way this is implemented first does anyone and not know what enumeration is
00:07:29
okay we're going to soothe everybody does does anybody then
00:07:33
it no ah everybody's and know why this is problematic
00:07:43
i couldn't quite has that sorry
00:07:47
that is a very good answer where to start um there's a
00:07:51
lot of problems that uh it mostly the whole thing is problematic
00:07:55
but the one thing i'm going to focus on 'cause i think it's pretty important is that
00:08:00
once you have written your admission like that where you
00:08:03
have declared integration of text acres with members okay not okay
00:08:08
pam this code is going to compile without any sort of warning
00:08:12
you can have a match on the value of type state stages the value and have a non exhaustive
00:08:18
that match is going to compile anything to fit at run time and we don't like it when things
00:08:23
for one side especially since the sky uh we tend to assume that
00:08:27
pattern match is smart and is going to want you whenever it's not exhaustive
00:08:32
this compound without a single wanting it crashes at run time it's everything that's wrong which was
00:08:38
a lot of years that we have is gonna so don't do that don't use enumeration um
00:08:44
used use cell trait hired is that uh uh well uh the way this works yeah it is
00:08:51
i have the greatest feel trait sealed means set set cannot be any subtype
00:08:56
of steak is declared outside of the file in which stages with like that
00:09:02
so the compiler knows that having a cat okay as a type of teachers not okay as the type of status
00:09:08
and those are the only sub types of data is in
00:09:11
this file that cannot possibly be any other subtype of status
00:09:16
so the component of now that this is not legal
00:09:21
then is going to warn you that you shouldn't be doing that anything interview
00:09:26
that is going to fail on that ideas we can give you quite a
00:09:30
good error message just because you stopped using enumeration so don't use animations ever
00:09:35
yeah just one thing this is actually a warning not a continuation failure
00:09:40
um up something that's pretty important if you want to write corrects gonna turn on fiddle warnings ah have
00:09:47
had a lot of but found that when i've never had a
00:09:50
false positive discusses very small that way warnings fixed and it's so important
00:09:58
um it's kind of free might eventually duplicate enumeration in favour of the in um q. it
00:10:04
uh have seen discussions on that matter last time i checked u. was
00:10:07
not the property gets that was not the proper poll request on that
00:10:11
uh but given the have check this weekend lots of things have been happening this week it might
00:10:15
actually be going on now i don't know but this is something that might actually disappeared from the language
00:10:22
uh_huh still trade so right i just told you so traits were nice and now
00:10:28
this is very similar to what i should do before but i
00:10:31
believe something is wrong with that code does anybody have and anyway
00:10:40
yes you in the back
00:10:44
i picked leather well thank you so i'm
00:10:50
at bar is not final and soprano yeah it's
00:10:55
when you see a seal traits and adjusted his definition sailed means you can't have a
00:11:00
subtype you can have a direct subtype outside of the filing with the type of the good
00:11:05
but direct is very important here most people don't really realise that but
00:11:11
but are is the director type of fruit and you can only have bar as
00:11:17
it directs the type of fruit anywhere in the rest of the code by itself so
00:11:21
yes the final so you can go to other file inductor a sector it you indicated some type of bile and
00:11:28
full by here is not a direct the type of food because as a level of indirection this bar in the middle
00:11:33
so this is correct this is perfect embedded this is exactly how feel that is supposed to behave
00:11:38
it's also for people who have just learned about sales not all that expect any kinda very surprising
00:11:45
so it's not should still and sometime it's what you might want to do but you have to
00:11:49
know that it behaves that way and unless you have a good reason for sub types of still trait
00:11:54
to be final to make them final or save themself a 'cause it's it's good default it's
00:11:59
not much which will what's a good default unless you have reasons to not follow that rule
00:12:06
uh right so solution is to prefer final still descendants and s. five
00:12:10
and can tell nothing's changing deserted that's it you know all dirty about that
00:12:16
um right
00:12:20
how to break that attack so i showed is that
00:12:22
just before uh it's basically the same code and now i'm
00:12:30
something's wrong with that something's where it was that yes you in the back
00:12:42
right so this is a very good answer but to the wrong slide i'm gonna very soon so can the answer and and and then and
00:12:50
say i think it's to slide down the road uh i'm not contributed because i don't want everybody to know the answers to the next trick
00:12:57
ah but no the problem here is start um this kinda compiler can sometime in for some
00:13:03
pretty wanted types so yeah if you create a list of digits okay instead just not okay
00:13:09
you'd expect the type to be list of status and instead you get this abomination
00:13:15
which doesn't we change all that much but again this is from the perspective of somebody learning the
00:13:21
language somebody who's new to the language in trying to work out how how how the body thing but
00:13:26
and when i see that the call help but think this is magic magic
00:13:30
objects and it's not what you'd expect and you don't understand what you have that
00:13:33
and the reason for that is perfectly correct coherent it's even possible for some people
00:13:38
to explain it i'm going to attempt at the actual wedding to get that but um
00:13:42
the reason this happens is that because here it's car is going to try and finds
00:13:48
the closest possible subtype uh is coco super type that is common to both okay not okay
00:13:54
one of these types the states because okay or not okay extend status but that's
00:13:59
not all 'cause by virtue of being cans objects okay not okay also extend implicitly
00:14:06
say reasonable and product which you don't need to know it's absolutely not interesting it was implementation detail
00:14:12
well it means that is the least this stuff about so the closest to the type that is
00:14:18
common to both of them is product with a reasonable we stages it's an implementation details that looks out
00:14:24
to the user doesn't wanna hear about that and who gets re compute
00:14:29
and when you have to explain that's what i just said to people who are learning schuyler
00:14:34
you lose about fifty percent of the yeah okay skies a bit magic at that time on that maybe we could make a backchannel script
00:14:40
so that doesn't really work and the way to trick the compiler into doing what you expect
00:14:46
is to have your entity extent product was veritable because of that and magically
00:14:53
okay do extend status and product it's a reasonable not okay as well but
00:14:58
both of the uh but stages is also a product that is readable and so
00:15:04
the list of abound is going to be status so this works this we had expected
00:15:10
sky freely enough improve on that um
00:15:14
i'm trying to hide it so i'm saying i'm keywords
00:15:19
whereby defaults not returned the type that you expect a return to super tight but it still
00:15:25
can happen if you use that if you use for constructors directly but i didn't have a lot
00:15:30
more interesting if you want to read about that uh this is clickable once i link to
00:15:35
slide at the end of the talk you can go and have a look at how this works
00:15:38
but um so dotty makes is that uh but it's not fixed is just kind of hidden under the rug
00:15:46
um there was that that was the answer so what's the problem is that case last
00:15:53
and uh as the answer was discussed as is non final and if it's
00:15:58
not final then get really weird behaviours that was in a sense you answer right
00:16:03
so an example of that behaviour is through being on final i'm not extended
00:16:10
here after class ball with two members i. n. s. which are in some
00:16:15
strings and i maps to choose i that's our problem and then we have
00:16:20
the additional string member and this means that this cold here is going to compile
00:16:25
and it's going to return true which is not what anybody said would expect
00:16:31
and i've actually taken the most obviously that's easiest example to
00:16:37
understand 'cause what happens if you put these two whoops sorry
00:16:42
and back there if you put these two in a hash map is that one of them is going to ride the
00:16:49
other and you're never going to figure out what and when that happens in production and you don't know why using elements um
00:16:57
it's happened to me i hope it doesn't happen to you it's a problem that now um
00:17:01
so the reason for that because i i promise that yeah that explain why this happens um
00:17:07
if some of you guys have some data experience what happens here is that through
00:17:13
uh being a case last gets a lot of synthetic code lots of methods generated for it such as
00:17:19
it falls to string hash code all these kind of things and so few
00:17:24
having equals method is that these implemented in a way that makes sense through basically wraps and and so if we
00:17:30
have to choose to have two things to have the same value them both of them are equal and that makes sense
00:17:34
but when you extend when you subtype ah through them bar inherits from the saint equals method
00:17:43
and so when you call that the method is going to be called is the method sars is a bit fiddly
00:17:51
the myth is equal method defining who which doesn't even know what s.
00:17:55
m. s. filled uh is it the string so this is absolutely normal
00:17:59
this is exactly what's expected when you know what's going on it's also
00:18:02
super confusing and and really does make the language for like magic so um
00:18:08
prefer making final case classes um uh the reason why this is not the default
00:18:14
the pen is is one person in the whole entire words that does because they're not for the case classes and has a use case from it
00:18:21
up that person inherently martin or the ski so when that happens you don't get to change it
00:18:26
uh i wish i wish to to change but unfortunately
00:18:30
that's nothing standing in these are two thirteen or dotty so
00:18:34
when you take a case last make it final please that's going to save all operating and
00:18:40
um custom instructors right does everybody know what a custom instructors
00:18:47
i'll take that as a yes um so what's going on here yeah is
00:18:53
and declaring an object extract some within and apply method for any psalm
00:18:59
is going to return and the value of type option of a and this allows me to use extract some of the pattern match and
00:19:06
this is really very long and does anybody know why when i found out
00:19:11
i i got kind of angry about that and i know what's wrong with that
00:19:18
nope okay have fun working it out so what happened say
00:19:23
when it is that this code here is going to compile
00:19:28
no problem even though it's quite clear that given the fact that my and apply method only texas some
00:19:36
it's not defined for non uh it's quite clear that hey if my though isn't on which can be
00:19:44
jesus i this is going to crash at run time because of how much does
00:19:50
not exhaustive but the compiler doesn't know that somehow and this will for run time
00:19:57
it's perfectly valid code by the compiler cannot work out that they have a problem so
00:20:04
so the reason for that was explained to me um it basically boils down to to really work that thing
00:20:11
uh i'm not construct explain it's a pattern matching exhaustively
00:20:14
checking is quite one key themes got to look twelve
00:20:18
uh and in order to avoid this kind of problem that solution would be
00:20:22
to never define custom x. factors if you can do that but if you can't
00:20:26
try to make the republic structures return some as an option because
00:20:31
that actually fixes the issue and you do get expected wanting here
00:20:36
it's kind of friendly make this whole thing overall patterns have been
00:20:40
a lot of research done on exhaustively t. checking for pattern matches
00:20:44
and uh so the point is completely gone in dirty and you don't need to worry about it as soon as you move just gonna free
00:20:50
but until then be careful was catholic structures they're weird
00:20:55
um structural types
00:20:59
do i need to give a quick quick information about what the structural type is does would anybody like that
00:21:05
think up of nods okay um structural typing is for those of you
00:21:10
who have used it languages this is doc typing popping a statically typed language
00:21:15
this allows me to say that my at one method here with accept any type
00:21:21
provided that type defines a i'm deaf uh get methods that resentment
00:21:27
and then i can just call that method 'cause you everything i know about that x. value is that
00:21:31
it's got a cat method so i get that in every time and at one that's what every time
00:21:36
um and then quite conveniently i can defend defined this case last notice how it's final important
00:21:43
um that rotten and and if i get my thought on that and i'm completely allowed
00:21:47
to pass wrapper to my i've been method even though i have not define any hard
00:21:53
some typing relationship so structural typing either way to do some typing without having to declare it beforehand
00:22:00
yeah it's quite convenient it's it's a great tool buttons colour he
00:22:03
suffers from a few falls does anybody know what the slows or
00:22:09
yes right okay so um the frenchman said reflection which basically covers my two point
00:22:17
um this is implemented in santa as reflections so the compiler knows this works
00:22:24
because it knows everything that isn't rubber it knows that rubber had a cat so it knows is called fail but at run time
00:22:31
is going to have to use reflection to dig into the values they
00:22:35
received extracts and make um a method object call that manually and this is
00:22:41
much more expensive than not doing anything so it has run time cost it's not nice even time cost and that should not
00:22:47
um discourage you from using structural typing we make sense but
00:22:52
the other problem is that i think some of the v. m.
00:22:56
this might not actually run at all 'cause it's possible and a during the day the m. to disable run time inflection
00:23:02
other configuration option for security managers so the composing to compare that there's no problem that code but
00:23:08
i wanted to find out that caught actually do that i have no way of knowing that in advance
00:23:15
uh was so the preferred approach is type classes and
00:23:22
i'm not having a bit over enthusiastic yeah so i
00:23:26
i don't think i have the time to control the slide if i go through this whole
00:23:28
thing um if anybody wants to know more about the type classes i'm going to find that
00:23:33
after the talk but i want to cover a few more of these before before we stopped
00:23:37
so i'm going to skip this nation if you know the type classes use that insubstantial type
00:23:43
uh and it's kind of freely them as overall z. and higher
00:23:48
federal type um it yeah i uh uh well i don't think it
00:23:53
changes of course the issue because the default implementations still relies
00:23:56
on reflection reflection and i don't see any ways that it
00:23:59
could be anything but reflection from the cases i don't think it's going to be fixed it's improved but not fix it all
00:24:08
exceptions
00:24:09
can you tell me what's wrong with these two bits of code
00:24:17
come again
00:24:20
right uh so pa since might throw number from an exception because most strings are not actually married into goes
00:24:26
and a lemon tarts might fro in the rain later that
00:24:29
exception or some such thing 'cause i might be outside of
00:24:34
he reached down and the problem with exceptions um well not very many
00:24:40
problems with exceptions but the one i've chosen to focus on here are
00:24:44
first thing is okay so this one is now and tiny
00:24:48
harnessed but it works wonders when trying to explain to other people
00:24:51
what happens very often is i'm working with a programmer redoing the code
00:24:54
that the buttons calla and when they getting by the input of one exception
00:24:58
and that point had that maybe you don't want from an exception maybe exceptions should not be used and they look at
00:25:03
me like i'm crazy i've been doing exceptions my whole life in java why should not do that what's wrong with them
00:25:09
and then what i do is ask them okay with the user go to segment in your code
00:25:14
course not nobody uses good to statements are famous papers written just about the fact that you should not be using go to statement
00:25:21
and then you ask that one what's the difference between good to stick with an exception can explain
00:25:25
to me what the critical difference between the two all in most of the time it happens they get
00:25:31
a blank for thirty seconds then so okay and they get rid of exceptionally carry on
00:25:37
so this is not an honest on my part because there's a very
00:25:40
many differences between exceptions and get two statements well there are few critical um
00:25:46
see now it is the first one is the problem with a good to statement is when
00:25:51
your at the label of a good to statement you have no idea where you coming from
00:25:55
you could come from anywhere it was good so you don't know your state and this is something that happens
00:26:00
exception as well you don't know was exceptions from you have like yeah you might have a double status that is
00:26:05
not gonna say could expect so you call no and this is very similar between good to statements exceptions
00:26:12
uh it's kind of in words for exceptions because when you write to go to statement to compare is going to
00:26:19
fail if there's no associated label so you know when you have to go to statement is going to go somewhere
00:26:25
with an exception is basically i'm going to pan out now so many
00:26:28
catch me otherwise crashed don't know don't care your problem and uh well
00:26:33
it unless you don't know that anybody's going to handle it you can't you can't take that so that's the
00:26:39
first reason why i don't like exceptions it's it's essentially good to statement for all the wrong things that now
00:26:45
um it's an checked it's gonna so this is basically just got a few other languages but in
00:26:51
some languages such as java exceptions are part of the type signature of the method that because i'm
00:26:57
so the compiler can tell you well you haven't caught the exception you haven't have that our case applies to
00:27:03
that and and of by the neck and compound but was colour the compiler complaining all exception you have no way
00:27:10
when you call method of for the exception you have no way to know whether you have
00:27:15
actually dealt with all possible error cases in that question one time and you have no way of taking that
00:27:20
and lastly east preferentially pack which is a very barbaric term which i'm going to explain now
00:27:29
so um referential transparent referential transparency um is
00:27:34
the property of code that we really like 'cause
00:27:36
it how obsessed reason reason about code at the local level evenings that uh when you have
00:27:43
an expression uh you can replace it by whatever it evaluates to without
00:27:49
changing the meaning of the code so you can ignore whatever else goes on
00:27:54
around the whole thing and focus on that expression that you all that matters so
00:27:57
expression is said to be refrain should be transparent if you can replay it put
00:28:03
what it's doing avoidable then use the mobile everywhere else in for an exception
00:28:08
if it was refreshing transparent would means that these two implementation would be the same as in in that one
00:28:14
i have a test that doesn't make much sense it's never going to be evaluated
00:28:18
well if falls it's true send and into for an exception of the budget and to
00:28:24
and the other implementation does the referential transparent testing where
00:28:29
it's going to put the result of that expression into horrible and then easy to use them the rest of the code
00:28:34
and obviously is that doesn't actually work out the first one with evaluating terminate
00:28:39
the second one fails so these two expressions are not equivalent and so
00:28:43
throwing exceptions is not refreshing transparent thank you don't want to do
00:28:47
that in your code 'cause it leads to all sorts of headaches
00:28:53
um you should it as much as possible prefer encoding your or types
00:28:58
in science of the return type of of the method to completely um
00:29:04
when frying a element but find exception if you want to convince the absence of you return an option
00:29:10
if you want to have some sort of fancy their code and all things like
00:29:14
that put it in an user so the right side of the users of successful part
00:29:18
the left side of these are these with the c. r. code the error message whatever
00:29:23
um and if you have if you absolutely have to
00:29:27
agree exceptions like you doing some general drop for example
00:29:30
try is going to help you use basically neither that contains an exception on
00:29:35
the left so if you have to be exceptions to try runs an uneasy
00:29:40
uh unfortunately nothing changed in dirty about that
00:29:46
now i'm starting to put together things that we have said before
00:29:50
uh uh this is one almost completely valid way of encoding or
00:29:57
so this is a very simplified entity which we talked about the fall um which
00:30:03
is meant to encode all possible error cases when you try to connect to database
00:30:07
and you know the for the whole thing to fit in one slide of just to get to possible implementation
00:30:12
one which is the sequel you try to run it not voted in the
00:30:15
other one is i try to connect but lost the connection please try again
00:30:20
and something is not ideal that implementation but it you know it doesn't have all the thing i've said so far
00:30:26
it is still trade extend product is a reasonable case objects everywhere it's it's it follows all the rules of
00:30:32
given so far but it's still not quite right or does anyone know why or how you could improve on that
00:30:41
okay out from that is that a lot of it yeah
00:30:45
it has to a library rely exceptions to cut the errors mostly
00:30:49
try and future and you have lots of libraries that we'll use future will use try
00:30:54
and you call stick your d. v. d. or enter into that if you said if you throw try
00:31:00
so when you know that work around that what you have to turn
00:31:03
out to do is use exceptions well not from them just typing it type
00:31:09
just typing it type is an exception
00:31:12
and so the the air is no exception was product answer usable
00:31:18
and you can absolutely stick values of type to be there in the
00:31:22
left or in the try sorry or in a future start problem at all
00:31:28
so the trick that but it allows your exception c. d. t.'s to
00:31:32
be a lot more usable when interacting with that discussed in a library
00:31:37
and right has returned to it um okay
00:31:43
uh what's wrong with that aside from everything
00:31:48
get off well down the return key word so there's lots and lots and lots
00:31:53
of reasons to not want to use it and um we has a whole go to
00:31:58
segment thing against basically same a discussion fracture friendship disparity again same discussion but mostly
00:32:04
is just completely insane ah it doesn't not at all do it i think it does
00:32:10
going to
00:32:13
well we're going to go for that right so this code yeah
00:32:19
ah looks kind of saying this is what somebody who's come from desperate for
00:32:24
example might write it makes a lot of sense ah you have a list of
00:32:27
names and you want to map on it and add one to absolutely was invite
00:32:31
you in it so you created an was function which returns and plus one i
00:32:36
unless you know to look for return this makes a lot of sense but the compiler hasn't quite agree
00:32:42
telling me that i have said it was a list of him but really should be an eight yeah
00:32:46
doesn't make much sense to me um mapping in the least should return a list but
00:32:50
compare it is often more smart than i am so i'm gonna just it and play along
00:32:55
so i do what he does when he suggests and now this
00:32:58
another list of in cement you complain should sorry should be stuff nothing
00:33:04
when nothing occur as in the type that you have inferred something's really wrong
00:33:10
um and i have no idea what's happening here but i'm just going to follow the error messages
00:33:15
back to it well you find a list of nothing but i really wanted him into what the hell
00:33:20
so i don't know what's going on i don't want to know what's going on i don't want to understand what's going on
00:33:25
i don't use return it makes no sense um you construct understand at some point you want you will fail you will
00:33:32
you will remember that slide and you will think i should not have flight understand it don't touch to it isn't that right
00:33:39
um so don't use return ever just from it um and set are
00:33:45
possible improvement in sky free in that non local written statement might be
00:33:50
forbidden i don't know what that means so maybe the thing i don't know
00:33:55
uh um it's implicit conversions right so this here methods defined as implicit
00:34:04
means that it's implicit conversions so whenever the compilers going to find a string
00:34:10
that he would like to be an aunt is going to call this method and turned my string in tonight
00:34:18
just looks pretty nice it's pretty it's pretty fun to you know what's wrong with that
00:34:26
right so if you muttering it i'm not going to hit a for that you came up with the answer
00:34:34
oh absolutely woeful an exception uh 'cause it lets you write beautiful go like
00:34:38
that full bar divided by two and a competitor is going to be very happy
00:34:42
to compare that and and carry on as if nothing happened intends is basically you lying to the compiler
00:34:48
the tank right i have as you do the string section and and
00:34:52
this is most of the time not choose there are infinitely many strings and infinitely many more strings that are not in
00:34:59
synching the or change so this is going to break eventually sooner rather than later and you never want to do that
00:35:06
the main problem being that this implicit conversion is not total
00:35:11
so if this was total so say transform in seem to float
00:35:16
it's actually but it's okay 'cause it called veteran tends to wear that you're getting the types like to
00:35:21
each other but it will not fit at run time but this is this is horribly don't wanna do that
00:35:26
uh instead well you could is is really nothing sorry you
00:35:35
could do this really nasty thing can be explicit about y'all
00:35:38
um conversions or you could just try not to divide for now is
00:35:42
to 'cause doesn't make much sense it's also possibility um so this is
00:35:49
at least it's got to that's a reason to happen to conversions is this is how will every that exposed to classes
00:35:57
will allow you to add syntax is is how you can have the map method when you have a factor instance for example is very nice
00:36:03
an implicit conversions to enrich existed types existing types are okay
00:36:08
the also completely and needed things kind of free because you have extension methods so
00:36:14
as soon as kind of free happens stop using data conversions the compiled into fighting over it anyway 'cause is going to ask you
00:36:20
to enable the computation flag and then to be very explicit about it uh it's
00:36:24
kind of free implicit conversions are kind of frowned upon and this is a good thing
00:36:32
implicit resolution so
00:36:37
i'm going to go for that one actually 'cause it's it's all um what happens here is
00:36:41
have to find the type config which contain the port for some sort of network configuration
00:36:47
and that if i get bored methods that for any config value in scope you can treat on the port so that way you could
00:36:55
implement a default configuration here so by default my configuration is going
00:36:59
to have port eighty eighty and if i called get port like that
00:37:04
then is going to work out that the default implementation here in your default pulled into the
00:37:09
eighty eighty it's quite common in the chasm but in skylight suffers from a pretty big problem
00:37:16
which has to do with uh impress each other wing so yeah in
00:37:20
another object i define a new implicit value of a completely different type
00:37:26
before coming here is that that config default config yeah of type int the only thing they have in common is their
00:37:35
name and that should not have any impact than the the name of the time when you result types but it does
00:37:43
we import both a one a two you get the dreaded cannot find good value for
00:37:47
parameter blah blah blah and this is very confusing because this is not what to expect
00:37:53
it's actually quite clear what happens when you remove simply sickness of
00:37:57
this if you were to import with a one a two and
00:38:00
ask the compiler to use the vocal fake easy so you what's ambiguous have
00:38:04
it both from a one and a two which one do you want to use
00:38:07
this is very clear this error message so when i don't have any
00:38:11
place it's thing stocking male anderson happens but what actually happens is you get
00:38:19
could not find peace that before parameter and this is black magic to most people who stop reading the language
00:38:26
this is extremely confusing and the bad news there is no real work abound
00:38:32
provided so that you don't have a perfectly unique name is going to happen eventually
00:38:37
and you can't guarantee that didn't have a computer you think anybody can be using the same as you and it was worse
00:38:44
thing is here when you do that you might not know that any to declare as an implicit value of the
00:38:49
wrong name so you might have no idea that you causing the problem you don't know what importing because importing everything
00:38:55
this is very confusing to beginners and what happens it's in that match explain it
00:39:00
so when you do it when you have to campus values given city names renal names
00:39:06
with with your name in its that nicholas is fancy full bar instant for type int
00:39:11
up it looks in it but the whole point it's an increase in value that is going to
00:39:14
see the name he's not supposed to maybe play the names of disembodied so so it's not a problem
00:39:19
but still not perfect somebody might also because nicholas and have the same in colleges i do so that my to be a problem
00:39:26
luckily in sky free first you have import implied which
00:39:30
means you knows that you imply importing implicit instinct is values
00:39:35
so it's less likely to happen because you call to do it by mistake but more importantly
00:39:39
um implicit danger in doesn't exist so the problem just completely gone it's gonna freeze it's major improvement
00:39:47
um ah string get nation
00:39:50
okay and anybody tell me what's wrong with that
00:39:58
good night
00:40:00
okay well um yeah string concatenation and using plus to distinguish generation this is this is a mistake because plus
00:40:09
this kind of insane it doesn't behave very coherently so yeah i have a
00:40:13
list of string and i plus bar it's or i dunno string to it
00:40:17
and deciding to return a string which has a value not super interesting was going to do that on the other hand if i have a set of strings
00:40:26
and i at but it's going to get to the collection and doesn't have the same type at all so this is with uh
00:40:34
when you know what happens and in if you understand it and you go ah
00:40:38
of course what that thinking but when try to expand that to somebody who's learning
00:40:41
the language you of course to behave that way coconut look at what the person's
00:40:45
going to look at lecture notes and he's going to go away never to sky anymore
00:40:49
uh and i'm not even talking about this absolute nightmare where you had to to list of
00:40:55
ins and the completion or is what i expected a string why why would expect a string
00:41:02
again it makes sense when you knows that any to string uh the the thing if you do know that these two two string ad
00:41:09
it does exist and what it does i'm sorry for your pain when fruit as well um up
00:41:14
i think it's the predicted um it is deprecated into that setting has been
00:41:20
removed from to that fourteen magically back in sky free i'm not sure why
00:41:25
uh i think it's going to i think is it just that's kind of free hasn't
00:41:28
been revenge on to that to that uh in standard library i hope i hope um but
00:41:35
so instead don't use a shouldn't get nation um use string interpretation instead
00:41:41
uh this is first i find a lot more readable then it doesn't have the same problem because it's using plus
00:41:48
and also yes they don't you know i was making the point that string customisation is much slower
00:41:53
than stricken depletion so that's that's well if you if you care about 'em efficiency use string information
00:42:01
uh in conclusion um
00:42:05
how do we enforce extract is is having lectures at all the stuff
00:42:10
that i have shown you and there's a lot more with it came from
00:42:13
i've i've just pick the best how do we make sure that this thing doesn't crop up in your code well
00:42:20
first could reduce i knows that a lot of people don't actually like could we use but there's some very solid research
00:42:27
that proves that code reviews actually does drastically reduce the amount of the facts in software
00:42:32
so when you get the slide you can click that he's going to link you to the publications that effect do code
00:42:38
reviews it's great and you have to avoid these things and you have to teach them to people who don't know them
00:42:44
use tools sets the to make the finding of these problems so that's a bunch of that's
00:42:51
i think so i think that at the moment that's all that exist but
00:42:55
let's get keep changing and we're happy valley beyond tuning at the moment so
00:42:58
i wouldn't be surprised that by next week we have another one but at
00:43:01
the moment gets can't go to what remove the scandal into its kind of style
00:43:06
it's gonna say especially if you go through rob notices pose a subject which uh makes kinda
00:43:12
kind of thing again and sky fix all these tools are going to make your code much much more reliable
00:43:18
also switched to scale officers can 'cause that you could see it fixed all of the problems that that that hard
00:43:24
um this well my favourite that um that well
00:43:31
either bad or best participating honestly look at it
00:43:34
um there's a lot more than them and i'm pretty sure i don't know all of them
00:43:38
so if you guys have things that you think i should have talked about and i haven't i'd
00:43:43
love to hear about that i'd i'd love to learn a bit more and i'd love to
00:43:46
include them into my next version of this tall going to the website upgraded to go with that
00:43:53
i i wanted to thank sees cannot contributors and dotty you
00:43:58
the channels they should earmarked i don't know whether he's here
00:44:00
uh but he's basically written all of c. e. a future improvements actions he's easy to be leaving database
00:44:07
we never asked something that but how does that work oh that's problem here publication here to get him
00:44:12
a depressed five second senior year of the hot i don't know he does that he's really quite impressive
00:44:17
well if you're wondering about sky that witnessing the language for how this one is
00:44:21
going to be fixing the future this to be the channels are really really great
00:44:26
a more informations so the slides are already
00:44:29
available here and um if you want to have
00:44:36
proper articles on all these rules and mall as as a companion site at this u. r. l.
00:44:42
well uh i to the time to to write these things down properly not just the slides
00:44:47
uh i can fully into the uh work they say though blah blah blah this is all done
00:44:51
with remote way remote yes and top hat on this problem is still here i sit in the fall
00:44:57
a tight is an amazing tool that will make sure that all the code you so in the slides is correct and evaluates to
00:45:03
what you've seen because it's actually executed when the slides i compiled
00:45:07
so it's not possible to gives like to compile and so the truth
00:45:11
and do we have any questions
00:45:15
oh
00:45:24
oh yes
00:45:30
yes thank you are thinking fast and i have a question rises colours needs and then uh
00:45:36
is it possible to and such cases what you described to me out
00:45:40
someone's been apart right so that's the point is it's made a tool so
00:45:45
was the first part of the question where is kinda fix 'cause right now what is where is what is what the question what is good
00:45:52
okay okay right uh and so these tools will automatically and
00:45:57
i the code and fragile built if they find on the distance
00:46:01
uh as a pretty opinionated what remover in particular is really hardcore with my
00:46:05
reports uh you can turn off some some things that you might not agree with
00:46:10
for example is that what remover is going to follow any code that defines default value for methods
00:46:15
but sometimes quite important to us and so i contacted with that ward so but
00:46:20
but all these tools will fan you build if they find some sort of pattern
00:46:25
and you can pick and choose an at whatever it wants and i mean i think that all the things i showed here are reported like one of these tools
00:46:34
i turned my back
00:46:47
oh i have a question about a using a try
00:46:51
ryan in other players learn except it may happen um
00:46:57
i have those experiences out there is such puerile um way
00:47:03
um my oh sure in all some slight orgy
00:47:07
oh oh so you'll write a cult and on
00:47:12
you learn to scold or in every place you should somehow handle
00:47:18
uh uh try um then should think about um
00:47:24
all the holes and you express dot membrane uh it will be some thresholds
00:47:30
not a okay in those cases i should eat user tried
00:47:34
but there are some cases slow or a a small it certainly is
00:47:40
completely okay right absolutely no this is this is a very
00:47:42
good question so exceptions are absolutely okay provided that actually are exceptional
00:47:49
so if you run into in areas such as a c. p. u. dot
00:47:52
files it's okay to crash you probably going to be able to recover from that
00:47:56
well uh the example i gave a word for example trying to do and and
00:48:00
number for that exception most strings are not things it's not exceptional case it's a
00:48:05
common case that this is a failure so this is not an exceptional exception it's
00:48:09
unexpected exception and you should not use exception for that the should use reason type
00:48:14
but for anything that you're not likely to supply from um or that just
00:48:19
doesn't make any sense uh just let that go absolutely let let's fatal errors
00:48:24
um be from that's on the problem at all i don't have much of a choice anyway i mean perhaps if you just if i'm not gonna do anyway
00:48:33
and for the question
00:48:36
don't matter whatever we don't have any posts are right okay so i don't have time properly
00:48:40
questions i will be here if you have other questions don't hesitate to come over and ask them
Conference Program
A Tour of Scala 3
Martin Odersky, Professor EPFL, Co-founder Lightbend
June 11, 2019 · 5:15 p.m.
8340 views
A story of unification: from Apache Spark to MLflow
Reynold Xin, Databricks
June 12, 2019 · 9:15 a.m.
1271 views
In Types We Trust
Bill Venners, Artima, Inc
June 12, 2019 · 10:15 a.m.
1569 views
Creating Native iOS and Android Apps in Scala without tears
Zahari Dichev, Bullet.io
June 12, 2019 · 10:16 a.m.
2233 views
Techniques for Teaching Scala
Noel Welsh, Inner Product and Underscore
June 12, 2019 · 10:17 a.m.
1296 views
Future-proofing Scala: the TASTY intermediate representation
Guillaume Martres, student at EPFL
June 12, 2019 · 10:18 a.m.
1159 views
Metals: rich code editing for Scala in VS Code, Vim, Emacs and beyond
Ólafur Páll Geirsson, Scala Center
June 12, 2019 · 11:15 a.m.
4697 views
Akka Streams to the Extreme
Heiko Seeberger, independent consultant
June 12, 2019 · 11:16 a.m.
1552 views
Scala First: Lessons from 3 student generations
Bjorn Regnell, Lund Univ., Sweden.
June 12, 2019 · 11:17 a.m.
577 views
Cellular Automata: How to become an artist with a few lines
Maciej Gorywoda, Wire, Berlin
June 12, 2019 · 11:18 a.m.
386 views
Why Netflix ❤'s Scala for Machine Learning
Jeremy Smith & Aish, Netflix
June 12, 2019 · 12:15 p.m.
5030 views
Massively Parallel Distributed Scala Compilation... And You!
Stu Hood, Twitter
June 12, 2019 · 12:16 p.m.
958 views
Polymorphism in Scala
Petra Bierleutgeb
June 12, 2019 · 12:17 p.m.
1113 views
sbt core concepts
Eugene Yokota, Scala Team at Lightbend
June 12, 2019 · 12:18 p.m.
1656 views
Double your performance: Scala's missing optimizing compiler
Li Haoyi, author Ammonite, Mill, FastParse, uPickle, and many more.
June 12, 2019 · 2:30 p.m.
839 views
Making Our Future Better
Viktor Klang, Lightbend
June 12, 2019 · 2:31 p.m.
1682 views
Testing in the postapocalyptic future
Daniel Westheide, INNOQ
June 12, 2019 · 2:32 p.m.
498 views
Context Buddy: the tool that knows your code better than you
Krzysztof Romanowski, sphere.it conference
June 12, 2019 · 2:33 p.m.
394 views
The Shape(less) of Type Class Derivation in Scala 3
Miles Sabin, Underscore Consulting
June 12, 2019 · 3:30 p.m.
2321 views
Refactor all the things!
Daniela Sfregola, organizer of the London Scala User Group meetup
June 12, 2019 · 3:31 p.m.
514 views
Integrating Developer Experiences - Build Server Protocol
Justin Kaeser, IntelliJ Scala
June 12, 2019 · 3:32 p.m.
551 views
Managing an Akka Cluster on Kubernetes
Markus Jura, MOIA
June 12, 2019 · 3:33 p.m.
736 views
Serverless Scala - Functions as SuperDuperMicroServices
Josh Suereth, Donna Malayeri & James Ward, Author of Scala In Depth; Google ; Google
June 12, 2019 · 4:45 p.m.
936 views
How are we going to migrate to Scala 3.0, aka Dotty?
Lukas Rytz, Lightbend
June 12, 2019 · 4:46 p.m.
710 views
Concurrent programming in 2019: Akka, Monix or ZIO?
Adam Warski, co-founders of SoftwareMill
June 12, 2019 · 4:47 p.m.
1975 views
ScalaJS and Typescript: an unlikely romance
Jeremy Hughes, Lightbend
June 12, 2019 · 4:48 p.m.
1377 views
Pure Functional Database Programming‚ without JDBC
Rob Norris
June 12, 2019 · 5:45 p.m.
6389 views
Why you need to be reviewing open source code
Gris Cuevas Zambrano & Holden Karau, Google Cloud;
June 12, 2019 · 5:46 p.m.
484 views
Develop seamless web services with Mu
Oli Makhasoeva, 47 Degrees
June 12, 2019 · 5:47 p.m.
785 views
Implementing the Scala 2.13 collections
Stefan Zeiger, Lightbend
June 12, 2019 · 5:48 p.m.
811 views
Introduction to day 2
June 13, 2019 · 9:10 a.m.
250 views
Sustaining open source digital infrastructure
Bogdan Vasilescu, Assistant Professor at Carnegie Mellon University's School of Computer Science, USA
June 13, 2019 · 9:16 a.m.
376 views
Building a Better Scala Community
Kelley Robinson, Developer Evangelist at Twilio
June 13, 2019 · 10:15 a.m.
245 views
Run Scala Faster with GraalVM on any Platform
Vojin Jovanovic, Oracle
June 13, 2019 · 10:16 a.m.
1342 views
ScalaClean - full program static analysis at scale
Rory Graves
June 13, 2019 · 10:17 a.m.
463 views
Flare & Lantern: Accelerators for Spark and Deep Learning
Tiark Rompf, Assistant Professor at Purdue University
June 13, 2019 · 10:18 a.m.
380 views
Metaprogramming in Dotty
Nicolas Stucki, Ph.D. student at LAMP
June 13, 2019 · 11:15 a.m.
1250 views
Fast, Simple Concurrency with Scala Native
Richard Whaling, data engineer based in Chicago
June 13, 2019 · 11:16 a.m.
624 views
Pick your number type with Spire
Denis Rosset, postdoctoral researcher at Perimeter Institute
June 13, 2019 · 11:17 a.m.
245 views
Scala.js and WebAssembly, a tale of the dangers of the sea
Sébastien Doeraene, Executive director of the Scala Center
June 13, 2019 · 11:18 a.m.
662 views
Performance tuning Twitter services with Graal and ML
Chris Thalinger, Twitter
June 13, 2019 · 12:15 p.m.
2003 views
Supporting the Scala Ecosystem: Stories from the Line
Justin Pihony, Lightbend
June 13, 2019 · 12:16 p.m.
163 views
Compiling to preserve our privacy
Manohar Jonnalagedda and Jakob Odersky, Inpher
June 13, 2019 · 12:17 p.m.
302 views
Building Scala with Bazel
Natan Silnitsky, wix.com
June 13, 2019 · 12:18 p.m.
565 views
Diversity and Inclusion Panel: Bring the Thunder!
Panel
June 13, 2019 · 2:30 p.m.
245 views
Asynchronous streams in direct style with and without macros
Philipp Haller, KTH Royal Institute of Technology in Stockholm
June 13, 2019 · 3:45 p.m.
304 views
Interactive Computing with Jupyter and Almond
Sören Brunk, USU Software AG
June 13, 2019 · 3:46 p.m.
682 views
Scala best practices I wish someone'd told me about
Nicolas Rinaudo, CTO of Besedo
June 13, 2019 · 3:47 p.m.
2713 views
High performance Privacy By Design using Matryoshka & Spark
Wiem Zine El Abidine and Olivier Girardot, Scala Backend Developer at MOIA / co-founder of Lateral Thoughts
June 13, 2019 · 3:48 p.m.
754 views
Immutable Sequential Maps – Keeping order while hashed
Odd Möller
June 13, 2019 · 4:45 p.m.
278 views
All the fancy things flexible dependency management can do
Alexandre Archambault, engineer at the Scala Center
June 13, 2019 · 4:46 p.m.
390 views
ScalaWebTest - integration testing made easy
Dani Rey, Unic AG
June 13, 2019 · 4:47 p.m.
468 views
Mellite: An Integrated Development Environment for Sound
Hanns Holger Rutz, Institute of Electronic Music and Acoustics (IEM), Graz
June 13, 2019 · 4:48 p.m.
213 views
Closing panel
Panel
June 13, 2019 · 5:54 p.m.
400 views