From b64fcfaa6115eb4a5f65c12d37f95d842136ae35 Mon Sep 17 00:00:00 2001
From: Adam Scarr <adam@vektah.net>
Date: Wed, 9 Aug 2017 22:14:27 +1000
Subject: [PATCH] Add a unicode whitespace parser

---
 parser.go      |  5 ++++-
 parser_test.go |  9 +++++++++
 state.go       | 50 +++++++++++++++++++++++++++++++-------------------
 3 files changed, 44 insertions(+), 20 deletions(-)

diff --git a/parser.go b/parser.go
index f084af4..7590ffa 100644
--- a/parser.go
+++ b/parser.go
@@ -25,6 +25,9 @@ type Result struct {
 //  - A parser that consumed some input should advance state.Pos
 type Parser func(*State) Result
 
+// VoidParser is a special type of parser that never returns anything but can still consume input
+type VoidParser func(*State)
+
 // Parserish types are any type that can be turned into a Parser by Parsify
 // These currently include *Parser and string literals.
 //
@@ -76,7 +79,7 @@ func ParsifyAll(parsers ...Parserish) []Parser {
 // WS will consume whitespace, it should only be needed when AutoWS is turned off
 func WS() Parser {
 	return NewParser("AutoWS", func(ps *State) Result {
-		ps.WS()
+		ps.WS(ps)
 		return Result{}
 	})
 }
diff --git a/parser_test.go b/parser_test.go
index 09cf7f4..bc2867e 100644
--- a/parser_test.go
+++ b/parser_test.go
@@ -203,6 +203,15 @@ func TestAutoWS(t *testing.T) {
 		require.Equal(t, "hello", result.Child[1].Token)
 		require.Equal(t, "", ps.Get())
 	})
+
+	t.Run("unicode whitespace", func(t *testing.T) {
+		ps := NewState(" \u202f hello")
+		ps.WS = UnicodeWhitespace
+
+		result := Exact("hello")(ps)
+		require.Equal(t, "hello", result.Token)
+		require.False(t, ps.Errored())
+	})
 }
 
 func runParser(input string, parser Parser) (Result, *State) {
diff --git a/state.go b/state.go
index 1860bc4..3c7de43 100644
--- a/state.go
+++ b/state.go
@@ -2,6 +2,8 @@ package goparsify
 
 import (
 	"fmt"
+	"unicode"
+	"unicode/utf8"
 )
 
 // Error represents a parse error. These will often be set, the parser will back up a little and
@@ -17,9 +19,6 @@ func (e Error) Pos() int { return e.pos }
 // Error satisfies the golang error interface
 func (e Error) Error() string { return fmt.Sprintf("offset %d: expected %s", e.pos, e.expected) }
 
-// WSFunc matches a byte and returns true if it is whitespace
-type WSFunc func(c byte) bool
-
 // State is the current parse state. It is entirely public because parsers are expected to mutate it during the parse.
 type State struct {
 	// The full input string
@@ -30,21 +29,41 @@ type State struct {
 	// in backtracking that it has been inlined to avoid allocations.
 	Error Error
 	// Called to determine what to ignore when WS is called, or when AutoWS fires
-	WSFunc   WSFunc
+	WS       VoidParser
 	NoAutoWS bool
 }
 
+// ASCIIWhitespace matches any of the standard whitespace characters. It is faster
+// than the UnicodeWhitespace parser as it does not need to decode unicode runes.
+func ASCIIWhitespace(s *State) {
+	for s.Pos < len(s.Input) {
+		switch s.Input[s.Pos] {
+		case '\t', '\n', '\v', '\f', '\r', ' ':
+			s.Pos++
+		default:
+			return
+		}
+	}
+}
+
+// UnicodeWhitespace matches any unicode space character. Its a little slower
+// than the ascii parser because it matches a rune at a time.
+func UnicodeWhitespace(s *State) {
+	for s.Pos < len(s.Input) {
+		r, w := utf8.DecodeRuneInString(s.Get())
+		if !unicode.IsSpace(r) {
+			return
+		}
+		s.Pos += w
+	}
+
+}
+
 // NewState creates a new State from a string
 func NewState(input string) *State {
 	return &State{
 		Input: input,
-		WSFunc: func(b byte) bool {
-			switch b {
-			case '\t', '\n', '\v', '\f', '\r', ' ':
-				return true
-			}
-			return false
-		},
+		WS:    ASCIIWhitespace,
 	}
 }
 
@@ -58,14 +77,7 @@ func (s *State) AutoWS() {
 	if s.NoAutoWS {
 		return
 	}
-	s.WS()
-}
-
-// WS consumes all whitespace and advances Pos.
-func (s *State) WS() {
-	for s.Pos < len(s.Input) && s.WSFunc(s.Input[s.Pos]) {
-		s.Pos++
-	}
+	s.WS(s)
 }
 
 // Get the remaining input.