go-parsekit/tokenhandler.go

package parsekit

import (
	"fmt"
)

// TokenHandler is the function type that is involved in turning a low level
// stream of UTF8 runes into parsing tokens. Its purpose is to check if input
// data matches some kind of pattern and to report back the match.
//
// A TokenHandler is to be used in conjunction with parsekit.P.On() or
// parsekit.Matcher().
//
// A TokenHandler function gets a TokenAPI as its input and returns a boolean to
// indicate whether or not it found a match on the input. The TokenAPI is used
// for retrieving input data to match against and for reporting back results.
type TokenHandler func(t *TokenAPI) bool

// TokenAPI is used by TokenHandler functions to retrieve runes from the
// input to match against and to report back results.
//
// Basic operation:
//
// To retrieve the next rune from the input, the TokenHandler function can call
// the TokenAPI.NextRune() method.
//
// The TokenHandler function can then evaluate the retrieved rune and either
// accept of skip the rune. When accepting it using TokenAPI.Accept(), the rune
//  is added to the resulting output of the TokenAPI. When using TokenAPI.Skip(),
// the rune will not be added to the output. It is mandatory for a TokenHandler
// to call either Accept() or Skip() after retrieving a rune, before calling
// NextRune() again.
//
// Eventually, the TokenHandler function must return a boolean value, indicating
// whether or not a match was found. When true, then the calling code will
// use the runes that were accepted into the TokenAPI's resulting output.
//
// Forking operation for easy lookahead support:
//
// Sometimes, a TokenHandler function must be able to perform a lookahead, which
// might either succeed or fail. In case of a failing lookahead, the state
// of the TokenAPI must be brought back to the original state.
//
// The way in which this is supported, is by forking a TokenAPI by calling
// TokenAPI.Fork(). This will return a child TokenAPI, with an empty
// output buffer, but using the same input cursor position as the forked parent.
//
// The TokenHandler function can then use the same interface as described for
// normal operation to retrieve runes from the input and to fill the resulting
// output. When the TokenHandler function decides that the lookahead was successful,
// then the method TokenAPI.Merge() can be called on the forked child to
// append the resulting output from the child to the parent's resulting output,
// and to update the parent input cursor position to that of the child.
//
// When the TokenHandler function decides that the lookahead was unsuccessful,
// then it can simply discard the forked child. The parent TokenAPI was never
// modified, so a new match can be safely started using that parent, as if the
// lookahead never happened.
type TokenAPI struct {
	p           *ParseAPI // parser state, used to retrieve input data to match against (TODO should be tiny interface)
	inputOffset int       // the byte offset into the input
	input       []rune    // a slice of runes that represents all retrieved input runes for the Matcher
	output      []rune    // a slice of runes that represents the accepted output runes for the Matcher
	currRune    *runeInfo // hold  information for the last rune that was read from the input
	parent      *TokenAPI // the parent MatchDialog, in case this one was forked
}

// runeInfo describes a single rune and its metadata.
type runeInfo struct {
	Rune     rune // an UTF8 rune
	ByteSize int  // the number of bytes in the rune
	OK       bool // false when the rune represents an invalid UTF8 rune or EOF
}

// NextRune retrieves the next rune from the input.
//
// It returns the rune and a boolean. The boolean will be false in case an
// invalid UTF8 rune or the end of the file was encountered.
//
// After using NextRune() to retrieve a rune, Accept() or Skip() can be called
// to respectively add the rune to the TokenAPI's resulting output or to
// fully ignore it. This way, a TokenHandler has full control over what runes are
// significant for the resulting output of that TokenHandler.
//
// After using NextRune(), this method can not be reinvoked, until the last read
// rune is explicitly accepted or skipped as described above.
func (t *TokenAPI) NextRune() (rune, bool) {
	if t.currRune != nil {
		panic("internal Matcher error: NextRune() was called without accepting or skipping the previously read rune")
	}
	r, w, ok := t.p.peek(t.inputOffset)
	t.currRune = &runeInfo{r, w, ok}
	if ok {
		t.input = append(t.input, r)
	}
	return r, ok
}

// Fork splits off a child TokenAPI, containing the same input cursor position
//  as the parent TokenAPI, but with all other data in a fresh state.
//
// By forking, a TokenHandler function can freely work with a TokenAPI, without
// affecting the parent TokenAPI. This is for example useful when the
// TokenHandler function must perform some form of lookahead.
//
// When a successful match was found, the TokenHandler function can call
// TokenAPI.Merge() on the forked child to have the resulting output added
// to the parent TokenAPI.
//
// When no match was found, the forked child can simply be discarded.
//
// Example case: A TokenHandler checks for a sequence of runes: 'a', 'b', 'c', 'd'.
// This is done in 4 steps and only after finishing all steps, the TokenHandler
// function can confirm a successful match. The TokenHandler function for this
// case could look like this (yes, it's naive, but it shows the point):
// TODO make proper tested example
//
//     func MatchAbcd(t *TokenAPI) bool {
//         child := t.Fork() // fork to keep m from input untouched
//         for _, letter := []rune {'a', 'b', 'c', 'd'} {
//             if r, ok := t.NextRune(); !ok || r != letter {
//                 return false // report mismatch, t is left untouched
//             }
//             child.Accept() // add rune to child output
//         }
//         child.Merge() // we have a match, add resulting output to parent
//         return true // and report the successful match
//     }
func (t *TokenAPI) Fork() *TokenAPI {
	return &TokenAPI{
		p:           t.p,
		inputOffset: t.inputOffset,
		parent:      t,
	}
}

// Accept will add the last rune as read by TokenAPI.NextRune() to the resulting
// output of the TokenAPI.
func (t *TokenAPI) Accept() {
	t.checkAllowedCall("Accept()")
	t.output = append(t.output, t.currRune.Rune)
	t.inputOffset += t.currRune.ByteSize
	t.currRune = nil
}

// Skip will ignore the last rune as read by NextRune().
func (t *TokenAPI) Skip() {
	t.checkAllowedCall("Skip()")
	t.inputOffset += t.currRune.ByteSize
	t.currRune = nil
}

func (t *TokenAPI) checkAllowedCall(name string) {
	if t.currRune == nil {
		panic(fmt.Sprintf("internal Matcher error: %s was called without a prior call to NextRune()", name))
	}
	if !t.currRune.OK {
		panic(fmt.Sprintf("internal Matcher error: %s was called, but prior call to NextRun() did not return OK (EOF or invalid rune)", name))
	}
}

// Merge merges the resulting output from a forked child TokenAPI back into
// its parent: The runes that are accepted in the child are added to the parent
// runes and the parent's input cursor position is advanced to the child's
// cursor position.
//
// After the merge, the child TokenAPI is reset so it can immediately be
// reused for performing another match (all data are cleared, except for the
// input offset which is kept at its current position).
func (t *TokenAPI) Merge() bool {
	if t.parent == nil {
		panic("internal parser error: Cannot call Merge a a non-forked MatchDialog")
	}
	t.parent.input = append(t.parent.input, t.input...)
	t.parent.output = append(t.parent.output, t.output...)
	t.parent.inputOffset = t.inputOffset
	t.ClearOutput()
	t.ClearInput()
	return true
}

// ClearOutput clears the resulting output for the TokenAPI, but it keeps
// the input and input offset as-is.
func (t *TokenAPI) ClearOutput() {
	t.output = []rune{}
}

// ClearInput clears the input for the TokenAPI, but it keeps the output
// and input offset as-is.
func (t *TokenAPI) ClearInput() {
	t.input = []rune{}
}