How is it even possible? In this article, I describe Scala performance tips based on my experience of optimizing . I found that some typical Scala code fragments have surprisingly much faster alternatives. The tips below can help you to avoid accidental performance issues and write faster Scala code for no additional effort. There is also source . Ergo's smart contracts interpreter code with all the benchmarks Scala provides powerful constructs and idioms to build abstractions for correctness and clarity but may have runtime performance penalties. When solving a task, there may be several alternative solutions, each with its own performance characteristics. Even similar-looking code can have radically different performance. Teams often specify things like formatting and documentation to make sure the code stays consistent and good quality. Similarly, tips on making the code run faster can also be useful. Performance improvement can come from almost no special effort. As such, it is a matter of coding style and habits rather than a matter of extensive code optimization, which shouldn't be . premature The code speedups described below are only examples and not necessarily exact numbers, as there may be various factors that can impact the timings. Nevertheless, after systematically applying the outlined recommendations, I have observed strictly positive effects on performance. Performance tips Tip 1: Avoid Seq() Let's start with the simplest. It's not uncommon to see the use of wherever an empty
collection is required. However, this method involves the following: Seq() an empty array is initiated the array is wrapped into an instance of ArraySeq the following methods called: -> -> -> SeqFactory$Delegate.apply IterableFactory.apply List.from Nil.prependAll Nil object returned in the end What to use instead Simple is 20-28% faster (see Benchmark: Seq) depending on whether the JIT is able to inline the invocation all the way down to or not. So, why not help it and use directly? Nil Seq() Nil Nil Tip 2: Avoid Map() Next one is . It is often used wherever an empty Map is required. However, as is the case with , this method involves allocations and many method calls. Map() Seq() What to use instead Calling is 25-85% faster, depending on the context (see  Benchmark: Map) Map.empty Tip 3: Avoid comprehensions for Looping pattern is used quite often due to its convenience.
It looks very similar to the loop in Java, but it is not. for (x <- xs) { ... } for In Scala, it is desugared by the compiler to which, besides execution of the block of code involves the following list of overheads: xs.foreach { x => ... } method call foreach allocation of lambda object boxing (memory allocation) of the value for every item (if xs values are not yet boxed) x xs hidden overhead of concrete implementation of method foreach Such comprehensions are particularly (see ) for Range loops, where is an . for sub-optimal Benchmark: for over Range x Int What to use instead There is a macro from library. cfor spire I recommend the following code as a replacement when provides an O(1) indexing access,
for example if is an or an array wrapped into . xs xs Array Seq import spire.syntax.cfor._
cfor(0)(_ < xs.length, _ + 1) { i => 
  val x = xs(i)
      ...
} The above collection can also be a in which case can be replaced with xs Range for (x <- 0 until n) { ... } cforRange(0 until n) { x => ... } These macros are compiled to efficient Java loop without overhead points 1) - 4). Depending on it is 10-170x faster (see Benchmark: for over Range). And since already implies a side effect operation, doesn't make the code any less readable. cfor xs.length foreach cfor Note, however, that is not a silver bullet. It is NOT a replacement for in the case of non-indexed collections, when indexed access is NOT a fast O(1) operation. In many cases using instead of is a no-brainer and often increases the performance of any code which loops over an indexed collection (see for example Benchmark: dot-product and Benchmark: check brackets). cfor foreach cfor foreach Tip 4: Avoid creating sequences with Seq(...) It is tempting to use method where a Seq of items is required like because it is easy and concise. You can pass it as a method argument or as a method result. Seq.apply Seq(1, 2, 3) However, the following happens under the hood: new Array with the items is created, and each item is boxed into its own Integer object 1, 2, 3 WrappedArray#ofInt wrapper is created and passed as vararg argument of Seq.apply the above-mentioned methods down to are called Nil.prependAll new ListBuffer is created and all items are copied from array to the buffer Benchmarks show that JIT is not able to fully optimize this code. What to use instead Simple drop-in replacement performs better. This code is 4-10x faster
than using . (see ). The speedup is even more significant when the number of items in the sequence grows (see ). Array(1, 2, 3) Seq(1, 2, 3) Benchmark: Seq vs Array Benchmark: Seq vs Array (from Range) What happens here is that: native unboxed Java array is created and then wrapped via implicit conversion into a which is inherited from the trait and can be used directly. WrappedArray#ofInt Seq Why is this faster? It avoids unnecessary allocations (only two new objects are allocated in the Array case).
Note that to create not only each Int is boxed into java.lang.Integer (or other primitive type),
but also instances are created for each item of the list. Seq scala.collection.immutable.:: Creating Array avoids boxing (which is proportional to the size of Seq). Accessing of array items is cache friendly both when the array is created and
when it is later used. Fewer allocations mean less garbage to collect later. This is especially important when the application is multithreaded because, in this case the garbage collector will compete with application threads for CPU resources, thus further slowing down the application. Notice that using arrays like this doesn't make the code any less readable. Moreover, the arrays created in this way are only accessible via interface, which is immutable. Thus, better performance is achieved without sacrificing other desirable code qualities like readability, safety, and conciseness. Seq Caveat: the array is not efficient if it is accessed using list operations and . See, for example, . In this case, replacing with may lead to performance degradation. However, using lists in a performance-critical code may not be a good idea as well. head tail Seq vs Array (list ops) benchmark Seq Array Tip 5: Avoid extractors with Option An extractor is an object with an method taking an arbitrary argument and returning an of some (maybe different) type. Here is a simple example: unapply Option object IsEven:
  def unapply(n: Int): Option[Int] =
    if (n % 2 == 0) Some(n) else None With extractors, pattern matching can be concise and expressive. However, returning an might have a negative impact on runtime performance because on every matching call an instance of needs to be created for each successful extraction. unapply Option Some What to use instead Since version 2.11, Scala introduces that no longer requires to return an . Combined with these two features allow implementing extractors, which are 10-70% faster than the original based extractors (see ). name based extractors unapply Option value classes allocation-free Option Benchmark: Option vs Opt object IsEvenOpt:
  def unapply(n: Int): Opt[Int] =
    if (n % 2 == 0) Opt(n) else Opt.empty A suitable implementation of class can be found in the library. Opt spire Benchmarks Summary The following tables summarize the results of the benchmarks run on the code snippets above and the following configuration: osArch: aarch64; osName: Mac OS X OpenJDK 64-Bit Server VM; vendor: Eclipse Adoptium; version: 17.0.5+8 You can reproduce the results on your computer by running the . Benchmarks class Clone the repository and run . sbt "Test/runMain benchmarks.Benchmark" Benchmark: Seq Size Seq(), ms Nil, ms Speedup 10 0.0001250 0.0001250 1.00 100 0.0002080 0.0002080 1.00 1000 0.0007500 0.0006250 1.20 10000 0.0065830 0.0051250 1.28 100000 0.0625830 0.0502920 1.24 Benchmark: Map Size Map(), ms Map.empty, ms Speedup 10 0.0001250 0.0001250 1.00 100 0.0002080 0.0001660 1.25 1000 0.0010410 0.0006250 1.67 10000 0.0091670 0.0050830 1.80 100000 0.0902500 0.0487080 1.85 Benchmark: foreach vs cfor Size foreach, ms cfor, ms Speedup 10 0.0006660 0.0000830 8.02 100 0.0007500 0.0000830 9.04 1000 0.0025830 0.0001250 20.66 10000 0.0238330 0.0003330 71.57 100000 0.2244170 0.0042910 52.30 Benchmark: for over Range Size foreach, ms cforRange, ms Speedup 10 0.0004160 0.0000830 5.01 100 0.0009160 0.0000830 11.04 1000 0.0061250 0.0001250 49.00 10000 0.0569170 0.0003330 170.92 100000 0.5648750 0.0042500 132.91 Benchmark: Seq vs Array (from elems) Size Seq, ms Array, ms Speedup 10 0.0007910 0.0002080 3.80 100 0.0066250 0.0008340 7.94 1000 0.0661660 0.0160000 4.14 10000 0.6733330 0.0679170 9.91 100000 6.8687920 0.6750420 10.18 Benchmark: Seq vs Array (from Range) Size Seq, ms Array, ms Speedup 10 0.0120420 0.0066250 1.82 100 0.0836670 0.0160830 5.20 1000 0.8315420 0.0533340 15.59 10000 8.1187080 0.5097500 15.93 100000 81.0221660 4.5335830 17.87 Benchmark: Seq vs Array (list ops) Size Seq, ms Array, ms Speedup 1 0.0003330 0.0003750 0.89 10 0.0005840 0.0007500 0.78 100 0.0030410 0.0042500 0.72 1000 0.0295000 0.1258330 0.23 10000 0.2695000 15.5551250 0.02 Benchmark: Option vs Opt Size Option, ms Opt, ms Speedup 100 0.0004160 0.0003750 1.11 1000 0.0014160 0.0009160 1.55 10000 0.0117910 0.0069580 1.69 100000 0.1155420 0.0666670 1.73 1000000 1.1547090 0.6657910 1.73 Benchmark: dot-product Size slow, ms fast, ms Speedup 10 0.0050000 0.0011250 4.44 100 0.0059160 0.0016660 3.55 1000 0.0242090 0.0073750 3.28 10000 0.1297920 0.1235000 1.05 100000 1.4926660 0.8446250 1.77 Benchmark: check brackets Size slow, ms fast, ms Speedup 1 0.0007080 0.0002500 2.83 10 0.0017080 0.0005000 3.42 100 0.0077500 0.0027090 2.86 1000 0.0753340 0.0260000 2.90 10000 0.7195830 0.2587080 2.78 Concluding remarks Program performance upkeep & optimization is the result of constant vigilance and work, rather than just a one-time job. There is no solution as there are
many trade-offs along the way. one size fits all The tips presented in this article are so simple that they can be applied and regardless of whether you have already identified the bottleneck or not. Back to my experience, once I systematically applied these tips all over the code base, those parts of code disappeared from the profiler's hot spots bringing up on the surface other bottlenecks by default and simplifies profiling. For further reading, I collected a list of useful references on the topic of Scala optimization. Happy optimizing! References Scala Style Guide Scala High Performance Programming (somewhat outdated) Optimizing Higher-Order Functions in Scala Where to look first when optimizing Scala code? Scala for comprehension performance Performance characteristics of Scala collections Java Performance: The Definitive Guide: Getting the Most Out of Your Code Scala library benchmarks JITWatch Parallel Collections: Measuring Performance JVM JIT optimization techniques Name Based Extractors in Scala 2.11

Sigma Protocols for the Working Programmer

Read My Stories

Too Long; Didn't Read

Make resilience your competitive advantage

How to Optimize Scala Code Performance With Almost No Effort

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

Bridging AI and Blockchain: Developing a Plugin for ChatGPT

06/09/2018: Biggest Stories in the Cryptosphere

07/09/2018: Biggest Stories in the Cryptosphere

0x Protocol, Waves, and EtherDelta: Solutions and Limitations of Distrust

03/09/2018: Biggest Stories in the Cryptosphere

Bridging AI and Blockchain: Developing a Plugin for ChatGPT

06/09/2018: Biggest Stories in the Cryptosphere

07/09/2018: Biggest Stories in the Cryptosphere

0x Protocol, Waves, and EtherDelta: Solutions and Limitations of Distrust

03/09/2018: Biggest Stories in the Cryptosphere

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps