I'm writing a parser in F#, and it needs to be as fast as possible (I'm hoping to parse a 100 MB file in less than a minute). As normal, it uses mutable variables to store the next available character and the next available token (i.e. both the lexer and the parser proper use one unit of lookahead).
My current partial implementation uses local variables for these. Since closure variables can't be mutable (anyone know the reason for this?) I've declared them as ref:
let rec read file includepath =
let c = ref ' '
let k = ref NONE
let sb = new StringBuilder()
use stream = File.OpenText file
let readc() =
c := stream.Read() |> char
// etc
I assume this has some overhead (not much, I know, but I'm trying for maximum speed here), and it's a little inelegant. The most obvious alternative would be to create a parser class object and have the mutable variables be fields in it. Does anyone know which is likely to be faster? Is there any consensus on which is considered better/more idiomatic style? Is there another option I'm missing?
You mentioned that local mutable values cannot be captured by a closure, so you need to use ref instead. The reason for this is that mutable values captured in the closure need to be allocated on the heap (because closure is allocated on the heap).
F# forces you to write this explicitly (using ref). In C# you can "capture mutable variable", but the compiler translates it to a field in a heap-allocated object behind the scene, so it will be on the heap anyway.
Summary is: If you want to use closures, mutable variables need to be allocated on the heap.
Now, regarding your code - your implementation uses ref, which creates a small object for every mutable variable that you're using. An alternative would be to create a single object with multiple mutable fields. Using records, you could write:
type ReadClosure = {
mutable c : char
mutable k : SomeType } // whatever type you use here
let rec read file includepath =
let state = { c = ' '; k = NONE }
// ...
let readc() =
state.c <- stream.Read() |> char
// etc...
This may be a bit more efficient, because you're allocating a single object instead of a few objects, but I don't expect the difference will be noticeable.
There is also one confusing thing about your code - the stream value will be disposed after the function read returns, so the call to stream.Read may be invalid (if you call readc after read completes).
let rec read file includepath =
let c = ref ' '
use stream = File.OpenText file
let readc() =
c := stream.Read() |> char
readc
let f = read a1 a2
f() // This would fail!
I'm not quite sure how you're actually using readc, but this may be a problem to think about. Also, if you're declaring it only as a helper closure, you could probably rewrite the code without closure (or write it explicitly using tail-recursion, which is translated to imperative loop with mutable variables) to avoid any allocations.
I did the following profiling:
let test() =
tic()
let mutable a = 0.0
for i=1 to 10 do
for j=1 to 10000000 do
a <- a + float j
toc("mutable")
let test2() =
tic()
let a = ref 0.0
for i=1 to 10 do
for j=1 to 10000000 do
a := !a + float j
toc("ref")
the average for mutable is 50ms, while ref 600ms. The performance difference is due to that mutable variables are in stack, while ref variables are in managed heap.
The relative difference is big. However, 10^8 times of access is a big number. And the total time is acceptable. So don't worry too much about the performance of ref variables. And remember:
Premature optimization is the root of
all evil.
My advice is you first finish your parser, then consider optimizing it. You won't know where the bottomneck is until you actually run the program. One good thing about F# is that its terse syntax and functional style well support code refactoring. Once the code is done, optimizing it would be convenient. Here's an profiling example.
Just another example, we use .net arrays everyday, which is also in managed heap:
let test3() =
tic()
let a = Array.create 1 0.0
for i=1 to 10 do
for j=1 to 10000000 do
a.[0] <- a.[0] + float j
toc("array")
test3() runs about the same as ref's. If you worry too much of variables in managed heap, then you won't use array anymore.
Related
I was reading Rust documentation section 4 and saw a code like this:
let mut s = String::from("hello");
let r1 = &mut s;
let r2 = &mut s;
println!("{}, {}", r1, r2);
So the documentation says that you cannot have multiple mutable references in Rust. Okay, makes sense but doc says three behaviors occurs if you could use, one of them is:
There’s no mechanism being used to synchronize access to the data.
Is there a need of mechanism to synchronize it? I mean we already use pointers to the heap or to another pointer that points to the heap.
I mean in this diagram, let's say that we have s2 and s3 as mutable references to s1.
s1 already has a pointer to the heap so s2 and s3 has pointers to s1. When we change s2 or s3 doesn't the memory change in the heap?
let mut s1 = String::from("Hello");
let s2 = &mut s1;
s2.push_str(", world");
In here the memory in the heap that s1 points to is changed so the s3 already points to that memory so doesn't it already synchronized?
I got the problem that why we should not use multiple mutable references. I just assume that we could. Rust says there is no mechanism to sync access to the data. My question is, we already have pointers to the heap from each reference so when we change the value in the heap it will all be synced because they are not values, they're just pointers to the heap and the value in the heap is changed?
Imagine if the following was allowed:
let mut maybe_five = Some(5);
let mut zero = 0;
// first mutable borrow
let ref_five = &mut maybe_five;
// second mutable borrow
// number_ref is a reference inside maybe_five's Some
let number_ref = match &mut maybe_five {
Some(ref mut five) => five,
_ => &mut zero,
};
// invalidate number_ref by making the value it points to invalid
*ref_five = None;
// now we write into a None???
*number_ref = 10;
We would be able to do all kinds of undefined behavior in safe rust.
I'm a little baffled about the inner work of the sequence expression in F#.
Normally if we make a sequential file reader with seq with no intentional caching of data
seq {
let mutable current = file.Read()
while current <> -1 do
yield current
}
We will end up with some weird behavior if we try to do some re-iterate or backtracking, My Idea of this was, since Read() is a function calling some mutable value we can't expect the output to be correct if we re-iterate. But then this behaves nicely even on boundary reading?
let Read path =
seq {
use fp = System.IO.File.OpenRead path
let buf = [| for _ in 0 .. 1024 -> 0uy |]
let mutable pos = 1
let mutable current = 0
while pos <> 0 do
if current = 0 then
pos <- fp.Read(buf, 0, 1024)
if pos > 0 && current < pos then
yield buf.[current]
current <- (current + 1) % 1024
}
let content = Read "some path"
We clearly use the same buffer to enhance performance, but assuming that we read the 1025 byte, it will trigger an update to the buffer, if we then try to read any byte with position < 1025 after we still get the correct output. How can that be and what are the difference?
Your question is a bit unclear, so I'll try to guess.
When you create a seq { }, you're essentially creating a state machine which will run only as far as it needs to. When you request the very first element from it, it'll start at the top and run until your first yield instruction. Then, when you request another value, it'll run from that point until the next yield, and so on.
Keep in mind that a seq { } produces an IEnumerable<'T>, which is like a "plan of execution". Each time you start to iterate the sequence (for example by calling Seq.head), a call to GetEnumerator is made behind the scenes, which causes a new IEnumerator<'T> to be created. It is the IEnumerator which does the actual providing of values. You can think of it in more classical terms as having an array over which you can iterate (an iterable or enumerable) and many pointers over that array, each of which are at different points in the array (many iterators or enumerators).
In your first code, file is most likely external to the seq block. This means that the file you are reading from is baked into the plan of execution; no matter how many times you start to iterate the sequence, you'll always be reading from the same file. This is obviously going to cause unpredictable behaviour.
However, in your second code, the file is opened as part of the seq block's definition. This means that you'll get a new file handle each time you iterate the sequence or, essentially, a new file handle per enumerator. The reason this code works is that you can't reverse an enumerator or iterate over it multiple times, not with a single thread at least.
(Now, if you were to manually get an enumerator and advance it over multiple threads, you'd probably run into problems very quickly. But that is a different topic.)
I've done most of my development in C# and am just learning F#. Here's what I want to do in C#:
string AddChars(char char1, char char2) => char1.ToString() + char2.ToString();
EDIT: added ToString() method to the C# example.
I want to write the same method in F# and I don't know how to do it other than this:
let addChars char1 char2 = Char.ToString(char1) + Char.ToString(char2)
Is there a way to add concatenate these chars into a string without converting both into strings first?
Sidenote:
I also have considered making a char array and converting that into a string, but that seems similarly wasteful.
let addChars (char1:char) (char2: char) = string([|char1; char2|])
As I said in my comment, your C# code is not going to do what you want ( i.e. concatenate the characters into a string). In C#, adding a char and a char will result in an int. The reason for this is because the char type doesn't define a + operator, so C# reverts to the nearest compatable type that does, which just happens to be int. (Source)
So to accomplish this behavior, you will need to do something similar to what you are already trying to do in F#:
char a = 'a';
char b = 'b';
// This is the wrong way to concatenate chars, because the
// chars will be treated as ints and the result will be 195.
Console.WriteLine(a + b);
// These are the correct ways to concatenate characters into
// a single string. The result of all of these will be "ab".
// The third way is the recommended way as it is concise and
// involves creating the fewest temporary objects.
Console.WriteLine(a.ToString() + b.ToString());
Console.WriteLine(Char.ToString(a) + Char.ToString(b));
Console.WriteLine(new String(new[] { a, b }));
(See https://dotnetfiddle.net/aEh1FI)
F# is the same way in that concatenating two or more chars doesn't result in a String. Unlike C#, it results instead in another char, but the process is the same - the char values are treated like int and added together, and the result is the char representation of the sum.
So really, the way to concatenate chars into a String in F# is what you already have, and is the direct translation of the C# equivalent:
let a = 'a'
let b = 'b'
// This is still the wrong way (prints 'Ã')
printfn "%O" (a + b)
// These are still the right ways (prints "ab")
printfn "%O" (a.ToString() + b.ToString())
printfn "%O" (Char.ToString(a) + Char.ToString(b))
printfn "%O" (String [| a;b |]) // This is still the best way
(See https://dotnetfiddle.net/ALwI3V)
The reason the "String from char array" approach is the best way is two-fold. First, it is the most concise, since you can see that that approach offers the shortest line of code in both languages (and the difference only increases as you add more and more chars together). And second, only one temporary object is created (the array) before the final String, whereas the other two methods involve making two separate temporary String objects to feed into the final result.
(Also, I'm not sure if it works this way as the String constructors are hidden in external sources, but I imagine that the array passed into the constructor would be used as the String's backing data, so it wouldn't end up getting wasted at all.) Strings are immutable, but using the passed array directly as the created String's backing data could result in a situation where a reference to the array could be held elsewhere in the program and jeopardize the String's immutability, so this speculation wouldn't fly in practice. (Credit: #CaringDev)
Another option you could do in F# that could be more idiomatic is to use the sprintf function to combine the two characters (Credit: #rmunn):
let a = 'a'
let b = 'b'
let s = sprintf "%c%c" a b
printfn "%O" s
// Prints "ab"
(See https://dotnetfiddle.net/Pp9Tee)
A note of warning about this method, however, is that it is almost certainly going to be much slower than any of the other three methods listed above. That's because instead of processing array or String data directly, sprintf is going to be performing more advanced formatting logic on the output. (I'm not in a position where I could benchmark this myself at the moment, but plugged into #TomasPetricek's benckmarking code below, I wouldn't be surprised if you got performance hits of 10x or more.)
This might not be a big deal as for a single conversion it will still be far faster than any end-user could possibly notice, but be careful if this is going to be used in any performance-critical code.
The answer by #Abion47 already lists all the possible sensible methods I can think of. If you are interested in performance, then you can run a quick experiment using the F# Interactive #time feature:
#time
open System
open System.Text
let a = 'a'
let b = 'b'
Comparing the three methods, the one with String [| a; b |] turns out to be about twice as fast as the methods involving ToString. In practice, that's probably not a big deal unless you are doing millions of such operations (as my experiment does), but it's an interesting fact to know:
// 432ms, 468ms, 472ms
for i in 0 .. 10000000 do
let s = a.ToString() + b.ToString()
ignore s
// 396ms 440ms, 458ms
for i in 0 .. 10000000 do
let s = Char.ToString(a) + Char.ToString(b)
ignore s
// 201ms, 171ms, 170ms
for i in 0 .. 10000000 do
let s = String [| a;b |]
ignore s
I'm just wondering if somebody can explain to me how to pass reference cells to functions that are not class members. I've been following the msdn page msdn reference cells
I have the following code:
let myint = ref 32
let mutable myint2 = 23
type addone() =
member t.myadd1func (x:int byref) =
x <- x + 1
let myadd1func (x:int byref) =
x <- x + 1
let adder = new addone()
adder.myadd1func myint
// myadd1func myint <---- this line does not compile
myadd1func &myint2 // <----- this line does though
printfn "%d" !myint
printfn "%d" myint2
My question is... what is the fundamental difference between the call I am making to the "Myadd1func" method on the class and the "myadd1func" function defined after it?
As I write this, I'm guessing that the function doesn't like having .net object references being passed to it as this might break compatibility with other IL components?? I don't mind using a mutable value, I just like to understand these things.
Thanks
I think the byref type in F# should be used only for interoperability purpsoes where the existing features (as explained by kvb) are good enough. If you want to declare a function that modifies some argument passed to it, I would just use ordinary reference cell (e.g. int ref type):
let myadd1func (x:int ref) =
x := !x + 1
let myint = ref 10
myadd1func myint
This may be slightly slower than using byref type (together with local mutable value), but I don't think it is needed very often in functional style, so it should be fine.
This is explained in the Type-directed Conversions at Member Invocations section of the F# specification. For interoperability with other .NET components, ref cells can be passed to members taking byref parameters and the compiler will automatically treat it as the dereferencing of the cell's contents field. However, this isn't done for let-bound functions, and you should directly use the addressof operator (&). You can still use a ref cell, but you have to explicitly dereference the contents field yourself, so this should work in your example: myadd1func &myint.contents
I have pieces of code like this in a project and I realize it's not
written in a functional way:
let data = Array.zeroCreate(3 + (int)firmwareVersions.Count * 27)
data.[0] <- 0x09uy //drcode
data.[1..2] <- firmwareVersionBytes //Number of firmware versions
let mutable index = 0
let loops = firmwareVersions.Count - 1
for i = 0 to loops do
let nameBytes = ASCIIEncoding.ASCII.GetBytes(firmwareVersions.[i].Name)
let timestampBytes = this.getTimeStampBytes firmwareVersions.[i].Timestamp
let sizeBytes = BitConverter.GetBytes(firmwareVersions.[i].Size) |> Array.rev
data.[index + 3 .. index + 10] <- nameBytes
data.[index + 11 .. index + 24] <- timestampBytes
data.[index + 25 .. index + 28] <- sizeBytes
data.[index + 29] <- firmwareVersions.[i].Status
index <- index + 27
firmwareVersions is a List which is part of a csharp library.
It has (and should not have) any knowledge of how it will be converted into
an array of bytes. I realize the code above is very non-functional, so I tried
changing it like this:
let headerData = Array.zeroCreate(3)
headerData.[0] <- 0x09uy
headerData.[1..2] <- firmwareVersionBytes
let getFirmwareVersionBytes (firmware : FirmwareVersion) =
let nameBytes = ASCIIEncoding.ASCII.GetBytes(firmware.Name)
let timestampBytes = this.getTimeStampBytes firmware.Timestamp
let sizeBytes = BitConverter.GetBytes(firmware.Size) |> Array.rev
Array.concat [nameBytes; timestampBytes; sizeBytes]
let data =
firmwareVersions.ToArray()
|> Array.map (fun f -> getFirmwareVersionBytes f)
|> Array.reduce (fun acc b -> Array.concat [acc; b])
let fullData = Array.concat [headerData;data]
So now I'm wondering if this is a better (more functional) way
to write the code. If so... why and what improvements should I make,
if not, why not and what should I do instead?
Suggestions, feedback, remarks?
Thank you
Update
Just wanted to add some more information.
This is part of some library that handles the data for a binary communication
protocol. The only upside I see of the first version of the code is that
people implementing the protocol in a different language (which is the case
in our situation as well) might get a better idea of how many bytes every
part takes up and where exactly they are located in the byte stream... just a remark.
(As not everybody understand english, but all our partners can read code)
I'd be inclined to inline everything because the whole program becomes so much shorter:
let fullData =
[|yield! [0x09uy; firmwareVersionBytes; firmwareVersionBytes]
for firmware in firmwareVersions do
yield! ASCIIEncoding.ASCII.GetBytes(firmware.Name)
yield! this.getTimeStampBytes firmware.Timestamp
yield! BitConverter.GetBytes(firmware.Size) |> Array.rev|]
If you want to convey the positions of the bytes, I'd put them in comments at the end of each line.
I like your first version better because the indexing gives a better picture of the offsets, which are an important piece of the problem (I assume). The imperative code features the byte offsets prominently, which might be important if your partners can't/don't read the documentation. The functional code emphasises sticking together structures, which would be OK if the byte offsets are not important enough to be mentioned in the documentation either.
Indexing is normally accidental complexity, in which case it should be avoided. For example, your first version's loop could be for firmwareVersion in firmwareVersion instead of for i = 0 to loops.
Also, like Brian says, using constants for the offsets would make the imperative version even more readable.
How often does the code run?
The advantage of 'array concatenation' is that it does make it easier to 'see' the logical portions. The disadvantage is that it creates a lot of garbage (allocating temporary arrays) and may also be slower if used in a tight loop.
Also, I think perhaps your "Array.reduce(...)" can just be "Array.concat".
Overall I prefer the first way (just create one huge array), though I would factor it differently to make the logic more apparent (e.g. have a named constant HEADER_SIZE, etc.).
While we're at it, I'd probably add some asserts to ensure that e.g. nameBytes has the expected length.