go-parsekit/parsekit.go

package parsekit

import (
	"fmt"
	"reflect"
	"runtime"
)

// Parser is the top-level struct that holds the configuration for a parser.
// The Parser can be instantiated using the parsekit.NewParser() method.
type Parser struct {
	startState StateHandler // the function that handles the very first state
}

// NewParser instantiates a new Parser.
//
// The Parser is a state machine-style recursive descent parser, in which
// StateHandler functions are used to move the state machine forward during
// parsing. This style of parser is typically used for parsing languages and
// structured data formats (like json, toml, etc.)
//
// To start parsing input data, use the method Parser.Parse().
func NewParser(startState StateHandler) *Parser {
	return &Parser{startState: startState}
}

// ParseRun represents a single parse run for a Parser.
type ParseRun struct {
	p *ParseAPI // holds the internal state of a parse run
}

// Parse starts a parse run on the provided input data.
// To retrieve emitted parser Items from the run, make use of the ParseRun.Next() method.
func (p *Parser) Parse(input string) *ParseRun {
	return &ParseRun{
		p: &ParseAPI{
			input:        input,
			len:          len(input),
			cursorLine:   1,
			cursorColumn: 1,
			nextState:    p.startState,
		},
	}
}

// Next retrieves the next parsed item for a parse run.
//
// When a valid item was found, then the boolean return parameter will be true.
// On error or when successfully reaching the end of the input, false is returned.
// When an error occurred, false will be returned and the error return value will
// be set (default is nil).
func (run *ParseRun) Next() (Item, *Error, bool) {
	// State handling loop: we handle states, until an Item is ready to be returned.
	for {
		// If a state handler has emitted one or more parser Items, then the next
		// available Item is returned to the caller.
		if len(run.p.items) > 0 {
			item, rest := run.p.items[0], run.p.items[1:]
			run.p.items = rest
			return run.makeReturnValues(item)
		}
		// Otherwise, the next state handler is looked up and invoked.
		run.runNextStateHandler()
	}
}

func (run *ParseRun) makeReturnValues(i Item) (Item, *Error, bool) {
	switch {
	case i.Type == ItemEOF:
		return i, nil, false
	case i.Type == ItemError:
		run.p.err = &Error{i.Value, run.p.cursorLine, run.p.cursorColumn}
		return i, run.p.err, false
	default:
		run.p.item = i
		return i, nil, true
	}
}

// runNextStateHandler moves the parser, which is bascially a state machine,
// to its next status. It does so by invoking a function of the
// type StateHandler. This function represents the current status and
// is responsible for moving the parser to its next status, depending
// on the parsed input data.
func (run *ParseRun) runNextStateHandler() {
	if state, ok := run.getNextStateHandler(); ok {
		run.invokeNextStateHandler(state)
	}
}

// getNextStateHandler determines the next StateHandler to invoke in order
// to move the parsing state machine one step further.
//
// When implementing a parser, the StateHandler functions must provide
// a routing decision in every invocation. A routing decision is one
// of the following:
//
// * A route is specified explicitly, which means that the next StateHandler
//   function to invoke is registered during the StateHandler function
//   invocation. For example: p.RouteTo(nextStatus)
//
// * A route is specified implicitly, which means that a previous StateHandler
//   invocation has registered the followup route for the current state.
//   For example: p.RouteTo(nextStatus).ThenTo(otherStatus)
//   In this example, the nextStatus StateHandler will not have to specify
//   a route explicitly, but otherStatus will be used implicitly after
//   the nextStatus function has returned.
//
// * An expectation is registered by the StateHandler.
//   For example: p.Expects("a cool thing")
//   When the StateHandler returns without having specified a route, this
//   expectation is used to generate an "unexpected input" error message.
//
// When no routing decision is provided by a StateHandler, then this is
// considered a bug in the state handler, and the parser will panic.
func (run *ParseRun) getNextStateHandler() (StateHandler, bool) {
	switch {
	case run.p.nextState != nil:
		return run.p.nextState, true
	case len(run.p.routeStack) > 0:
		return run.p.popRoute(), true
	case run.p.expecting != "":
		run.p.UnexpectedInput()
		return nil, false
	default:
		name := runtime.FuncForPC(reflect.ValueOf(run.p.state).Pointer()).Name()
		panic(fmt.Sprintf("internal parser error: StateHandler %s did not provide a routing decision", name))
	}
}

// invokeNextStateHandler moves the parser state to the provided state
// and invokes the StateHandler function.
func (run *ParseRun) invokeNextStateHandler(state StateHandler) {
	run.p.state = state
	run.p.nextState = nil
	run.p.expecting = ""
	run.p.state(run.p)
}

// Matcher is the top-level struct that holds the configuration for
// a parser that is based solely on a TokenHandler function.
// The Matcher can be instantiated using the parsekit.NewMatcher()
// method.
//
// To match input data against the wrapped Matcher function, use the method
// Matcher.Parse().
type Matcher struct {
	parser *Parser
}

// NewMatcher instantiates a new Matcher.
//
// This is a simple wrapper around a TokenHandler function. It can be used to
// match an input string against that TokenHandler function and retrieve the
// results in a straight forward way.
//
// The 'expects' parameter is used for creating an error message in case parsed
// input does not match the TokenHandler.
func NewMatcher(tokenHandler TokenHandler, expects string) *Matcher {
	stateHandler := func(p *ParseAPI) {
		p.Expects(expects)
		if p.On(tokenHandler).Accept() {
			p.EmitLiteral(0) // ItemType is irrelevant
		}
	}
	return &Matcher{parser: NewParser(stateHandler)}
}

// Parse checks for a match on the provided input data.
func (m *Matcher) Parse(input string) (string, *Error, bool) {
	item, err, ok := m.parser.Parse(input).Next()
	if !ok {
		return "", err, false
	}
	return item.Value, nil, true
}