Disclaimer!
This post is a little bit ridiculous and over the top but I'd be doing injustice if I do not share my thoughts on how the journey of learning something new feels like. But if you're in a rush, just scroll to the snippet at the bottom
So here's the thing. Years ago, when I had just started programming, as part of my degree, I hated C#. I didn't want anything to do with any project related to Microsoft related tech. Here's the rationale behind this (it's a dumb rationale) Why does C# have a 'C' in its name but looks almost exactly like Java? I also believed that a good programmer must know Java, C/C++, PHP for server programming, and Python just to prove you're smart. I swear this is a phase that I literally went through.
This was because when you've just started coding, for some reason there's this notion of being defined by the programming languages you know and the libraries you use. I'm sure that was greatly influenced by the "Top 10 best languages to learn in 201X" or "Language X is dead, here's why", even "Programming languages used by Facebook" articles that I actively looked for! just so that I don't get left behind in the industry.
Of course, a lot changed in the way I view tech and my career but what I felt about C# or anything related to it never changed, until about 6-7 months ago. See, at work, they use a shit ton of .NET (or dotnet. I'm still getting the hang of this) so I'm learning it now. A few months ago, I completed a 'Basics for beginners' udemy course. It was fine I guess๐ but learned a lot of cool stuff ๐ค
Now I'm creating a project so that I learn more of this freaky stuff. Enter .NET Core 3! The most overwhelming piece of framework I've ever encountered (this year. I have to be clear on this). I have never given up so many times but come back to try again, in just a single week, ever for any project that I ever started to learn. When I finally got my thingie to give me a token for authentication, after days and nights of struggle, I ran the most satisfyingly powerful git command known to mankind
git commit -m 'Initial commit'
I was happy. Then I had my supper while rereading up on what I've implemented and that's when I saw it
The masterpiece
private User GetUserByToken(string token)
{
var user = (from u in _context.Users
join t in _context.Tokens
on u.Id equals t.UserId
where t.Body == token
select u).SingleOrDefault();
return user;
}
It's beautiful isn't it? I thought so too
I mean do you understand that the last time I saw something this cool was when I saw JSX for the first time, which looked so confusing but really not so trivial after some time of use. My brain doesn't have to context switch between application code and database scripts but the thing just works seamlessly. And the syntax highlighting plus the IntelliSense! I am stunned.
If you're reading this and you do not understand WTF is going on here, it's okay, it's not your time yet. But later on in your life, you will see what I saw and you'll say "shit, that dude on dev.to wrote a weird post about this feeling I'm feeling right now".
Top comments (64)
I hate C# LINQ query syntax. It is completely pointless. You can always use LINQ methods directly.
This is a warning to anyone who thinks it's cool.
It gives you nothing. It's not shorter, it's not more readable, it's not faster... It's actually harder to refactor and maintain.
I used to love them at first: they look so cool, right? Well, I've been using C# for many, many years, and the experience has always converged to: avoid them, you will regret the temptation. - Use lambdas directly instead: Lambdas are great.
Nice simple example: Often, you need to split the query into 2 places: super simple with normal syntax: you just copy part of it elsewhere, keeping it as IEnumerable/IQueryable. - With LINQ syntax, you now have to figure out how to split the nonlinear query, the beginning and end is always repeated, you have to type them over..... Zero added value, only problems. After all, its design was inspired by the ergonomically ~worst language: SQL. (I admit, this is opinionated; what I say about LINQ syntax: not so much)
Most importantly: Not all LINQ/extension methods have LINQ syntax, so now you have to mix it, and it's completely horrible.
Also, it has nothing to do with .NET Core: It's been in C# since before 2008.
btw. Don't get me wrong: LINQ extension methods are great... except, for no good reason, C# decided to call all the methods differently than every language ever - but the LINQ syntax was 100% a mistake.
I love how this came after I had seen this comment firsthttps://dev.to/fernandomondo/comment/ilae or else I would have been devastated by now. I have rewriten that piece of code in my project using lambda, that whole function is made up of just two lines now. In fact, this is a great neede eye-opener.
The LINQ syntax has some advantages in certain cases. It can be both more readable and shorter if your query is complex (sub-queries/grouping/cross products) or requires temporary state (those tend to not be translatable to SQL and are usually done in memory).
Lambdas add a ton of noise in general because of punctuation (parentheses, braces, arrows) and you have to re-declare the item variable in every chained call.
For a simple query like this using a navigation property and a lambda a simpler, as shown.
It is about knowing when and how to use a tool; most tools are not categorically "useless".
@maartyl here goes me proving you wrong! ๐
But with this argument, we all come out as winners. The abundance of tools means there's more than one way to do the job.
@brunnerh Thank you for a counterexample. I agree that with complex queries they probably become more worthwhile, and at some point almost certainly are nicer than using lambdas directly, however, I would argue a complex query should not even be kept in the code, usually.
(I should note, I'm not too familiar with LINQ to SQL)
It can probably be refactored into multiple reusable functions. (and the moment you do this, the LINQ syntax will be worse again) - It probably should be, and using LINQ syntax makes it harder to do so. - At least to me, large queries seem quite hard to read and reason about, whereas 'composing function{s, calls}' is much nicer (I admit, this is probably a matter of opinion, but I often reuse 'subqueries' and that is generally better).
(if it is LINQ to objects) I've never seen that complex queries actually needed - Usually, there is no problem using 'lambda' syntax. (although, to be fair: I haven't always tried, but the trying would probably cost more time overall, than just always using one pattern for everything - especially one that INTERRACTS well with everything else in the language)
(if it is LINQ to SQL) If it is complex enough: it should probably not be written in C# at all, as LINQ syntax is very limiting, and I've repeatedly found it does not support things I needed, and had to rewrite it in raw SQL anyway.
Writing a query that mixes SQL and memory (does not translate to SQL) feels like a code smell to me. I understand there is some benefit to having a purely SQL query, which you then change to actually run in memory, without rewriting it, but it feels more like it would happen by accident, and not behaving as you expect it to. - I like to have a clear separation of what runs where, and generally clear separation between all things that need their own 'governing'.
Lambda noise was never much of an issue for me, but I admit this is a good point. I use lambdas everywhere for everything, but someone not as used to them might find it annoying.
Overall, there probably is some use for them (for actual, complex queries) but in my experience, I don't think I've ever encountered it. I guess they may be great for someone else, I just cannot imagine it...
PS: No idea why dev.to notified me the first time, but not again. I've only found this because I came back to report on my own findings.
I'm glad you saw the other comment first. ^^ I would have hated to make you feel devastated. I just wanted to deter others from repeating my mistakes. ^^ (and clearly, I'm not the only one :D) - I was probably just a bit harsher, because that temptress burnt me many times, before I finally learned. XD
PS: I would love it, if you could prove me wrong. ^^
- I would love to use them... it's just never been worth it.
This is going to be hard ๐ You have years of experience, I have a few weeks. Unless they introduce a breaking change that we'll most likely learn at the same pace, there's no way I can defend LINQ as of yet
I want to wholeheartedly thank you for this article. I considered LINQ syntax so useless, I forgot about it when trying to figure out how to bend C# to my needs. Thanks to you, I started to wonder what LINQ actually could be useful for, and it has a use!
It can be used to add monadic 'do' syntax sugar. I missed this for a long time, creating all sorts of workarounds. Inventing ways to use the language in manners the language was never meant to be used. (Not stopping now, lol)
Now, I finally found a use for LINQ syntax. I didn't implement it yet, but I am nearly certain it's possible. Implementing the specific SelectMany on my custom type, I should be able to write 'nested' bind calls (thus able to reference previous variables) as a flat sequence of operations.
(I just hope there is not some horrible trap, like there is in mutable async blocks (which makes them useless as generic monad syntax))
Imagine
m()
to be some monadic expression.If this works, my code will be so much more readable and maintainable thanks to this. ^^ - And if it wasn't for you, I would have never thought about this. So, thank you. ^^
I am a fan of the fluent API. I never much liked the 'SQL-like' variants. Joins especially were ugly. Getting back an IQueryable and working with that till you need the data is much preferred, IMO.
Here is your example written using an async method and lambda expression syntax:
private async Task<User> GetUserByTokenAsync(string token) => await _context.Users
.Join(_context.Tokens.Where(t => t.Body == token), u => u.Id, t => t.UserId, (u, t) => u)
.SingleOrDefaultAsync();
Note: you will need to include additional namespaces for the async extension methods on IQueryable<> and for Task<>
Drop the async and await keywords for better performance.
At some point, I'm looking forward to benchmarking the alternatives i.e. Fluent vs Lambda, Async vs Sync. I'm certain there's no one answer for every use case
Fluent translates to the same underlying Linq statements and concequently the same generated SQL. Note: The await keyword can be used in front of the parentheses.
Async code frees up the thread to do other things while it's waiting for the slower database network roundtrip. You are trading a small setup cost for starting the state machine handling the async processing, but that's it.
So basically async statements are useful for performance in situations where the data being processed is rather large
I agree that using the Async flavor of the SingleOrDefault is generally the preferred way as it improves massively the amount of load a single server can absorb. My comment was simply that the proposed method does not need to be async (contain the keyword async in the signature) as there is no need to await tasks and execute continuations within the function. So, basically, you can directly return the Task returned by SingleOrDefaultAsync without awaiting it as the caller will await it.
@Beautus S Gumede on "So basically async statements are useful for performance in situations where the data being processed is rather large"
(Not seeing proper "Reply" function on mobile)
That's one example where async is advantageous, and there are many, and it gets complicated pretty quick. I would recomend doing a deep dive on Asynchronous and Threading, but know that they are two very different subjects (although their impacts on your code can appear similar).
There's a pretty common analogy used to explain threading involving a chef/cook that I find to be a good foundation for internalizing the concept. Examples are all over Google for that. This article gives you a reasonable overview with some good detail, IMO: medium.com/swift-india/concurrency...
Happy coding!
@Beautus S Gumede In a highly performant scalable web app, async/await allows .NET to momentarily give the underlying thread to a different incoming request while waiting on IO (such as a database call). It makes .NET prolly the most scalable stack out there when done right. NodeJS does similar things with its singlethreaded event loop.
This is much better understandable I think
Great! Now try lambda expression, you will be happier:)
I did ๐ it looks way better if you're working on a single table
You can use navigation properties and lambda to improve your code.
Something like:
public class User
{
public virtual Token Token { get; set; }
}
And then:
_context.Users.SingleOfDefault(u => u.Token.Body == Token);
Ok, this actually looks cleaner and more concise. I'm assuming from this that the virtual property relation to users is setup automatically because both entities exist in one context. I'm puzzled by how this really works
Here's another strange thing I'm picking up from this snippet, the fact that the property name and its type can just be exactly the same without C# complaining. I've been experimenting areas where such is possible.
public virtual Token Token;
I will add my voice to the others here who recommend avoiding LINQ syntax. Most MS development shops have gotten away from LINQ syntax and your query shows one good reason:
It might be more efficient for the
where
to execute before thejoin
. With the LINQ syntax you can do it but it looks odd.Now compare:
For your specific case, I suggest you turn
_context.Tokens
into anIDictionary<string, Token>
or evenIDictionary<string, User>
where the token body is the key. Then your method looks like this:I have a question with this one. How would my DBContext class look like, will this UsersByToken field have to be mapped to a table in the bd too?
One thing that has been mentioned a couple of times is that I should consider using IDictionary, I'm looking forward to it
Ah. I did not know what
_context
was and I was not aware you were using EF. I assumed_context
was some kind of property bag or class you controlled. So, the answer is no you would not have aUsersByToken
property on yourDbContext
and it would not map to the database. Instead it should be a local field/property in your login class. Now that I understand you are making round trips to the database, you will findIDictionary<>
to be even more efficient than your current code, but you will need a fall back to retrieving the user from the database if it is not found associated the provided token in your dictionary.Thank you for this. Seems like IDictionary and IQueryable are my best bet for getting the most out of EF
You shouldn't use the join keyword, a second "from" is best:
i) you get to use the equality operator;
ii) you can arrange the operands any way you like;
iii) you can add on the DefaultIfEmpty() function on the end to change to a left join and vice versa.
Also, I always use the LINQ format instead of chained functions. Simply because it's more readable.
This is so true. The join was unnecessary ๐ค I was kind of suprised to why the Lambda version I switched to seems shorter, it was because it didn't use a join ๐ฌ
Yeah... LINQ is solid stuff. Had a dev on my team give the same reaction a few weeks back.
Now if you want it to run 10 times faster, drop entity framework and use dapper
If you want better performance but still strongly typed you could look at linq2db
The next step up from this is to combine this with webapi odata, that way the framework generated this sort of linq expression for you from a URL.
The irony here is the discovery .net core 3 and the EF core libs that make this possible are right now in a worse state than the ef6 and .net 4.x stuff regular .net people just migrated from.
Give it a couple years though and this will be special under .net 5 when you start notice it's literally write once, run anywhere.
LINQ, like SQL, is declarative.
It's interesting that I didn't' use the term 'declarative' but now that you've mentioned it, everything makes sense. I like declarative code :D
Roll with it. Read 'T-SQL Fundamentals, Third Edition' by Itzik Ben-Gan. Work through all the examples. You'll see why "SELECT" really belongs after "FROM" and "WHERE" (LINQ knows!). Find a good LINQ resource and work through more examples (LINQ to this, LINQ to that...). Don't worry about the lambda expressions underpinning LINQ syntactic sugar at this time. Microsoft has a fine offering. Visual Studio is the best IDE on the market. DOT NET "Core" is platform independent. For Web, perhaps check out angular.io (a Google framework) underpinned by TypeScript language (the "C# guy" invented TypeScript). Angular is organized, TypeScript is awesome, and it all downcompiles into the correct JavaScript for the device requesting your "stuff." There are dollar bills for your bank account on the Microsoft train. A lot of the other stuff is budget busting, shifting sand. In my experience.
I 100% agree that SQL got select wrong, and it should be at the end!
As for VS being the best IDE, I honestly find using Kotlin and IntelliJ Idea more pleasant.
I felt it bro, C# is easy and beautiful