nasin/config/file.go

package config

import "io"
import "fmt"
import "math"
import "bufio"
import "errors"
import "strconv"
import "strings"
import "unicode"

type line any
type comment string
type entry struct { key string; value Value }

// File represents a config file. It preserves the order of the lines, as well
// as blank lines and comments.
type File struct {
	lines []line
	keys  map[string] int
}

// NewFile creates a blank file with nothing in it.
func NewFile () *File {
	return &File {
		 keys: make(map[string] int),
	}
}

// Parse parses a config file from a reader. This function operates on a
// best-effort basis: A file will always be returned, and any errors encountered
// will be joined together.
//
// The general format of the file is as follows:
//   - Encoded in UTF-8
//   - Consists of lines, separated by \n, or \r\n
//   - Lines can be any of these:
//     - Blank line: has only whitespace
//     - Comment: begins with a '#'
//     - Entry: a key/value pair separated by an '=' sign
//
// For entries, all whitespace on either side of the '=' sign, the key, or the
// value is ignored. The key may contain any letter or digit, as well as '-'
// and '.'. The value is always identified by its first rune (after the
// preliminary whitespace of course) and can be one of:
//   - String: either a double-quoted string, or any string of runes not
//     identifiable as any other kind of value. The quoted string is always
//     unquoted when it is read. Either way, these escape sequences are
//     supported, and resolved when they are read:
//     - '\\': a literal backslash
//     - '\a': alert, bell
//     - '\b': backspace
//     - '\t': horizontal tab
//     - '\n': line feed
//     - '\v': vertical tab
//     - '\f': form feed
//     - '\r': carriage return
//     - '\"': double quote
//   - Number: a floating point value. It can be of the form:
//     - Inf
//     - -Inf
//     - NaN
//     - [0-9]+
//     - [0-9]+\.[0-9]*
//   - Bool: a boolean value. It can be one of:
//     - true
//     - false
//
// Be aware that some unquoted strings, within reason, are subject to being read
// as some other value in the future. For example, if there were suddenly a
// third boolean value called glorble, the unquoted string glorble would be read
// as a boolean value instead of a string.
func Parse (reader io.Reader) (*File, error) {
	file := &File {
		keys: make(map[string] int),
	}
	errs    := []error { }
	scanner := bufio.NewScanner(reader)
	
	for scanner.Scan() {
		text := strings.TrimSpace(scanner.Text())
		switch {
		case text == "",  strings.HasPrefix(text, "#"):
			file.lines = append(file.lines, comment(text))
		default:
			entry, err := parseEntry(text)
			if err == nil {
				file.keys[entry.key] = len(file.lines)
				file.lines = append(file.lines, entry)
			} else {
				errs = append (errs, err )
			}
		}
	}
	errs = append(errs, scanner.Err())
	return file, errors.Join(errs...)
}

func parseEntry (str string) (entry, error) {
	key, value, ok := strings.Cut(str, "=")
	if !ok { return entry { }, ErrMalformedEntry }
	
	key   = strings.TrimSpace(key)
	if !KeyValid(key) { return entry { }, ErrMalformedKey }

	value = strings.TrimSpace(value)
	parsedValue, err := ParseValue(value)
	if err != nil { return entry { }, err }
	
	return entry { key: key, value: parsedValue }, nil
}

// ParseValue parses a value from a string. For any Value v,
// ParseValue(v.String()) should hold data exactly equal to v. This function
// does not trim whitespace.
func ParseValue (str string) (Value, error) {
	first := firstRune(str)
	switch {
	case str == "":
		return ValueString(""), nil
	case first == '"':
		value, ok := unescape(str)
		if !ok { return nil, ErrMalformedEscapeSequence }
		value, ok = unquote(value)
		if !ok { return nil, ErrMalformedStringValue }
		return ValueString(value), nil
	case first == '-' || (first >= '0' && first <= '9'):
		value, err := strconv.ParseFloat(str, 64)
		if err != nil { return nil, ErrMalformedNumberValue }
		return ValueNumber(value), nil
	case str == "false":
		return ValueBool(false), nil
	case str == "true":
		return ValueBool(true), nil
	case str == "Inf":
		return ValueNumber(math.Inf(1)), nil
	case str == "NaN":
		return ValueNumber(math.NaN()), nil
	default:
		value, ok := unescape(str)
		if !ok { return nil, ErrMalformedEscapeSequence }
		return ValueString(value), nil
	}
}

// Has returns whether the key exists. If the key is invalid, it returns false,
// ErrMalformedKey.
func (this *File) Has (key string) (bool, error) {
	if !KeyValid(key) { return false, ErrMalformedKey }
	_, ok := this.keys[key]
	return ok, nil
}

// Get gets the keyed value. If the value is unspecified, it returns nil,
// ErrNonexistentEntry. If the key is invalid, it returns nil, ErrMalformedKey.
func (this *File) Get (key string) (Value, error) {
	if !KeyValid(key) { return nil, ErrMalformedKey }
	if index, ok := this.keys[key]; ok {
		if lin, ok := this.lines[index].(entry); ok {
			return lin.value, nil
		}
	}
	return nil, ErrNonexistentEntry
}

// Set sets a value. If the key is invalid, it returns ErrMalformedKey.
func (this *File) Set (key string, value Value) error {
	if !KeyValid(key) { return ErrMalformedKey }
	ent := entry {
		key:   key,
		value: value,
	}
	if index, ok := this.keys[key]; ok {
		this.lines[index] = ent
		return nil
	}
	this.keys[key] = len(this.lines)
	this.lines     = append(this.lines, ent)
	return nil
}

// Reset removes the value from the file. If the value is set again, it will be
// added back at the same location. Note that because of this, the positions of
// lines are not forgotten until the file is written and reloaded. This is why
// the method is called Reset and not Remove. If the key is invalid, it returns
// ErrMalformedKey.
func (this *File) Reset (key string) error {
	if !KeyValid(key) { return ErrMalformedKey }
	for index, lin := range this.lines {
		if lin, ok := lin.(entry); ok {
			if lin.key == key {
				this.lines[index] = nil
			}
		}
	}
	return nil
}

// Map creates and returns a map of keys to values.
func (this *File) Map () map[string] Value {
	mp := make(map[string] Value)
	for key, index := range this.keys {
		if lin, ok := this.lines[index].(entry); ok {
			mp[key] = lin.value
		}
	}
	return mp
}

// Diff returns a set of keys that are different from the other file.
func (this *File) Diff (other *File) map[string] struct { } {
	diff := make(map[string] struct { })
	
	// - keys only we have
	// - keys we both have, but are different
	for key, index := range this.keys {
		otherIndex, ok := other.keys[key]
		if !ok {
			diff[key] = struct { } { }
			continue
		}
		if !this.lines[index].(entry).value.Equals(other.lines[otherIndex].(entry).value) {
			diff[key] = struct { } { }
		}
	}

	// - keys only they have
	for key := range other.keys {
		if _, has := this.keys[key]; !has {
			diff[key] = struct { } { }
		}
	}

	return diff
}

// WriteTo writes the data in this file to an io.Writer.
func (file *File) WriteTo (writer io.Writer) (n int64, err error) {
	for _, lin := range file.lines {
		nint := 0
		switch lin := lin.(type) {
		case comment:
			nint, err = fmt.Fprintln(writer, string(lin))
		case entry:
			nint, err = fmt.Fprintf(writer, "%s=%v\n", lin.key, lin.value)
		}
		n += int64(nint)
		if err != nil { return n, err }
	}
	return n, nil
}

// KeyValid returns whether a key contains only valid runes. They are:
//   - Letters
//   - Digits
//   - '-'
//   - '.'
func KeyValid (key string) bool {
	for _, char := range key {
		valid :=
			char == '.' ||
			char == '-' ||
			unicode.IsLetter(char) ||
			unicode.IsDigit(char)
		if !valid { return false }
	}
	return true
}