Pattern matching numeric strings - f#

I have a function that pattern matches its argument, which is a string:
let processLexime lexime
match lexime with
| "abc" -> ...
| "bar" -> ...
| "cat" -> ...
| _ -> ...
This works as expected. However, I'm now trying to extend this by expressing "match a string containing only the following characters". In my specific example, I want anything containing only digits to be matched.
My question is, how can I express this in F#? I'd prefer to do this without any libraries such as FParsec, since I'm mainly doing this for learning purposes.

You can use active patterns: https://msdn.microsoft.com/en-us/library/dd233248.aspx
let (|Integer|_|) (str: string) =
let mutable intvalue = 0
if System.Int32.TryParse(str, &intvalue) then Some(intvalue)
else None
let parseNumeric str =
match str with
| Integer i -> printfn "%d : Integer" i
| _ -> printfn "%s : Not matched." str

One way would be an active pattern
let (|Digits|_|) (s:string) =
s.ToCharArray() |> Array.forall (fun c -> System.Char.IsDigit(c)) |> function |true -> Some(s) |false -> None
then you can do
match "1" with
|Digits(t) -> printf "matched"

I would use regular expressions combined with active patterns. With regular expressions you can easily match digits with \d and active patterns makes the syntax nice inside your match.
open System.Text.RegularExpressions
let (|ParseRegex|_|) regex str =
let m = Regex("^"+regex+"$").Match(str)
if (m.Success) then Some true else None
let Printmatch s =
match s with
| ParseRegex "w+" d -> printfn "only w"
| ParseRegex "(w+|s+)+" d -> printfn "only w and s"
| ParseRegex "\d+" d -> printfn "only digis"
|_ -> printfn "wrong"
[<EntryPoint>]
let main argv =
Printmatch "www"
Printmatch "ssswwswwws"
Printmatch "134554"
Printmatch "1dwd3ddwwd"
0
which prints
only w
only w and s
only digis
wrong

Related

How can I use TryParse in a guard expression in match?

I have built a toy spreadsheet to help learn F#. When I process the text for a new cell I store it as a discriminated type. To parse it I feel I should be able to do something like:
let cv =
match t with
| _ when t.Length=0 -> Empty
| x when t.[0]='=' -> Expr(x)
| x when t.[0]='\"' -> Str(x)
| (true,i) when Int32.TryParse t -> IntValue(i) // nope!
| _ -> Str(t)
I have tried quite a few combinations but I cannot get TryParse in the guard. I have written a helper:
let isInt (s:string) =
let mutable m:Int64 = 0L
let (b,m) = Int64.TryParse s
b
I can now write:
| _ when Utils.isInt t -> IntValue((int)t)
This seems like a poor solution as it discards the converted result. What the correct syntax to get TryParse into the guard?
I think an active pattern will do what you want:
let (|Integer|_|) (str: string) =
let flag, i = Int32.TryParse(str)
if flag then Some i
else None
let cv =
match t with
| _ when t.Length=0 -> Empty
| x when t.[0]='=' -> Expr(x)
| x when t.[0]='\"' -> Str(x)
| Integer i -> IntValue(i)
| _ -> Str(t)
But if you really want TryParse in the guard condition (and you don't mind parsing twice), you could do this:
| x when fst (Int32.TryParse(t)) -> IntValue (Int32.Parse(x))

Pattern matching against List variable

Why does the second pattern matching return "1 and 2" and give me a warning "This rule will never be matched" on the wildcard _?
let a = [3;4]
match a with
|[1;2] -> "1 and 2"
|_ -> "Other"
|> printfn "%A"
let lst = [1;2]
match a with
|lst -> "1 and 2"
|_ -> "Other"
|> printfn "%A"
It appears to me lst is considered truthy. How so?
An identifier in lowercase matches with everything and binds the identifier to the value.
If you want to compare lst with a you need to use when which is a conditional guard:
let a = [3;4]
match a with
|[1;2] -> "1 and 2"
|_ -> "Other"
|> printfn "%A"
let lst = [1;2]
match a with
| b when b = lst -> "1 and 2"
|_ -> "Other"
|> printfn "%A"
but in this case a simple if then else would work.
In your 2nd pattern matching you have two different lst variables with the same name. Here,
let lst = [1;2]
match a with
| lst -> "1 and 2"
| _ -> "Other"
in the 1st match case you don't refer to the lst variable above, you create a new one. You can check it by trying to refactor-rename it (F2 in Visual Studio). As AMieres explains, case is important.
In general, referring to a variable in a match case is doable yet not trivial.
Consider this code:
let helloWorld = "hello world"
let isHelloWorld s =
match s with
| helloWorld -> true
| _ -> false
Here you will get the same warning as described. One way to go is to mark the constant with the [<Literal>] attribute and to make it upper-case at the same time:
[<Literal>]
let HelloWorld = "hello world"
let isHelloWorld s =
match s with
| HelloWorld -> true
| _ -> false
This would work as expected. However, you can apply the [<Literal>] attribute only to certain types, and list is not among them. For those, you have to leverage when conditions in match cases.

How can I remove a character from a string while doing character by character processing?

I'm fairly new to F# but I'm struggling to find how to properly represent the null character in the language. Can anyone tell me how to represent the null character in F#?
More to the point, what started me down the path is I'm trying to do some string processing with String.mapi, but I can't figure out how to remove a character in the below function:
let GetTargetFrameworkFolder version =
let versionMapper i c =
match c with
| 'v' -> if i = 0 then char(0x000) else c
| '.' -> char(0x000)
| _ -> c
match version with
| "v3.5" -> "net35"
| "v4.0" -> "net40"
| "v4.5" -> "net45"
| vers -> vers |> String.mapi versionMapper
GetTargetFrameworkFolder "v4.5.1" |> Dump
How can I remove a character from a string while doing character by character processing, as in the case with String.map and String.mapi?
You cannot remove a character using String.mapi, as this function maps exactly one character from the input to one character from the output. The null character is not the same thing as removing a character; it's just another character that happens to have the code 0.
In your case, if I understand correctly you want to remove the initial 'v' (if any) and remove dots. I would do it like this:
let GetTargetFrameworkFolder version =
match version with
| "v3.5" -> "net35"
| "v4.0" -> "net40"
| "v4.5" -> "net45"
| vers ->
let vers = if vers.[0] = 'v' then vers.[1..] else vers
vers.Replace(".", "")
Another way of doing this if you wanted to keep your original approach would be to write your own choose function for strings:
module String =
let choosei predicate str =
let sb = System.Text.StringBuilder()
let choose i (c:char) =
match predicate i c with
| Some(x) -> sb.Append(c) |> ignore
| None -> ()
str |> String.iteri choose
sb.ToString()
Then use it as follows:
let GetTargetFrameworkFolder version =
let versionMapper i = function
| 'v' when i = 0 -> None
| '.' -> None
| c -> Some(c)
match version with
| "v3.5" -> "net35"
| "v4.0" -> "net40"
| "v4.5" -> "net45"
| vers -> vers |> String.choosei versionMapper
GetTargetFrameworkFolder "v4.5.1" |> Dump
You can achieve this by using an array comprehension:
let GetTargetFrameworkFolder version =
match version with
| "v3.5" -> "net35"
| "v4.0" -> "net40"
| "v4.5" -> "net45"
| vers -> new String([|
for i in 0 .. vers.Length - 1 do
match i, vers.[i] with
| 0, 'v' | _, '.' -> () // skip 'v' at [0] and all '.'s
| _, c -> yield c // let everything else through
|])
By character processing while removing a character is filtering (string is a sequence of char):
let version (s: String) =
s
|> Seq.filter (fun ch -> ch <> '.' && ch <> 'v')
|> String.Concat
UPDATE:
To skip first 'v':
let version (s: String) =
s
|> Seq.skip (if s.StartsWith "v" then 1 else 0)
|> Seq.filter ((<>) '.')
|> String.Concat

Create Discriminated Union Case from String

I'm trying to create DU cases from strings. The only way I can see doing this is by enumerating over the DU cases via Microsoft.FSharp.Reflection.FSharpType.GetUnionCases and then picking the UnionCase that matches the string (by using .Name) and then making the actual DU case out of that by using FSharpValue.MakeUnion.
Isn't there an easier/more elegant way of doing this? In my scenario I have a DU with a couple of hundred cases for keywords. I have to read the strings (keywords) from a file and make the types out of them. I did some "optimization" by putting the cases into a Map but I was hoping there'd be a better way of doing this.
I have the following, for example:
type Keyword =
| FOO
| BAR
| BAZ
| BLAH
let mkKeywords (file: string) =
use sr = new StreamReader(file)
let caseMap =
FSharpType.GetUnionCases(typeof<Keyword>)
|> Array.map (fun c -> (c.Name, FSharpValue.MakeUnion(c, [||]) :?> Keyword))
|> Map.ofArray
[
while not sr.EndOfStream do
let l = sr.ReadLine().Trim()
match caseMap.TryFind l with
| Some c -> yield c
| None -> failwith <| "Could not find keyword: " + l
]
I found this handy code snippet...
open Microsoft.FSharp.Reflection
let toString (x:'a) =
let (case, _ ) = FSharpValue.GetUnionFields(x, typeof<'a>)
case.Name
let fromString<'a> (s:string) =
match FSharpType.GetUnionCases typeof<'a> |> Array.filter (fun case -> case.Name = s) with
|[|case|] -> Some(FSharpValue.MakeUnion(case,[||]) :?> 'a)
|_ -> None
... which makes it easy to tack on two lines of code to any DU...
type A = X|Y|Z with
override this.ToString() = FSharpUtils.toString this
static member fromString s = FSharpUtils.fromString<A> s
I would use pattern matching like this:
type Keyword =
| FOO
| BAR
| BAZ
| BLAH
let matchKeyword (word:string) : Keyword option =
match word with
| "FOO" -> Some FOO
| "BAR" -> Some BAR
| "BAZ" -> Some BAZ
| "BLAH" -> Some BLAH
| _ -> None
And maybe auto generate the match statement first time using regex in my editor, but only because you have hundreds of cases. But i am not sure if its a better solution then yours.
As the cases have no value, another option is to use enums:
type Keyword =
| FOO = 0
| BAR = 1
| BAZ = 2
| BLAH = 3
let strings = ["FOO";"BAR"]
let keywords =
[for s in strings -> s, Keyword.Parse(typeof<Keyword>, s)]
|> Map.ofList
Then you can simply use Enum.Parse.

Literal Attribute not working

After reading Chris' answer to F# - public literal and the blog post at http://blogs.msdn.com/b/chrsmith/archive/2008/10/03/f-zen-the-literal-attribute.aspx I don't get why the following is not working:
[<Literal>]
let one = 1
[<Literal>]
let two = 2
let trymatch x =
match x with
| one -> printfn "%A" one
| two -> printfn "%A" two
| _ -> printfn "none"
trymatch 3
This keeps printing "3", although I think it shouldn't. What is it that I don't see here?
I think that literals need to be Uppercase. The following works fine:
[<Literal>]
let One = 1
[<Literal>]
let Two = 2
let trymatch x =
match x with
| One -> printfn "%A" One
| Two -> printfn "%A" Two
| _ -> printfn "none"
trymatch 3
In addition, if you want a nice general solution for this without using literals, you can define a parameterized active pattern like this:
let (|Equals|_|) expected actual =
if actual = expected then Some() else None
And then just write
let one = 1
let two = 2
let trymatch x =
match x with
| Equals one -> printfn "%A" one
| Equals two -> printfn "%A" two
| _ -> printfn "none"
The other answers are right - you must start your identifier with an uppercase letter. See section 7.1.2 of the spec (Named Patterns), which states that:
If long-ident is a single identifier that does not begin with an uppercase character then it is always interpreted as a variable-binding pattern and represents a variable that is bound by the pattern
Also if you don't want to have Uppercase literals you can put them in a module (here named Const):
module Const =
[<Literal>]
let one = 1
[<Literal>]
let two = 2
let trymatch x =
match x with
| Const.one -> printfn "%A" Const.one
| Const.two -> printfn "%A" Const.two
| _ -> printfn "none"
trymatch 3
Don't ask me why, but it works when you write your literals uppercase:
[<Literal>]
let One = 1
[<Literal>]
let Two = 2
let trymatch (x:int) =
match x with
| One -> printfn "%A" One
| Two -> printfn "%A" Two
| _ -> printfn "none"
trymatch 3

Resources