DateKeys/internal/pathrule/rules.go

package pathrule

import (
	"fmt"
	"regexp"
	"strings"
	"unicode/utf8"
)

// Limits of the paths and texts of a format 3 head, fixed with the format
// (spec §29.4, §29.5).
const (
	MaxPathLen        = 1024  // R1, bytes
	MaxSegments       = 32    // R2
	MaxSegmentLen     = 255   // R3, bytes
	MaxSegmentUTF16   = 255   // R3, UTF-16 code units of the NFD of the segment
	MaxImplicitDirs   = 65535 // R9
	MaxCommentLen     = 16384 // bytes
	MaxAuthorLen      = 256   // bytes
	reservedDirPrefix = ".datekeys-"
)

// Error is the violation of one rule of spec §29.5 or §29.6. Its text is
// the same in every implementation, so that the message of a rejected head
// does not depend on the reader.
type Error struct {
	Rule   string // "R2" to "R10", "text"
	Detail string
	// Paths are the positions, from 1, of the two paths of an R7
	// violation, the later one first, so that a writer can name them; zero
	// for the other rules.
	Paths [2]int
}

func (e *Error) Error() string { return e.Rule + ": " + e.Detail }

func fail(rule, format string, args ...any) error {
	return &Error{Rule: rule, Detail: fmt.Sprintf(format, args...)}
}

// codePoint formats r as in the spec: U+ and at least four upper-case
// hexadecimal digits.
func codePoint(r rune) string { return fmt.Sprintf("U+%04X", r) }

// forbiddenASCII are the ASCII characters of R4 that are not controls.
const forbiddenASCII = `"*:<>?\|`

// CheckPath checks one path of the head with the rules of the fourth layer
// that concern it alone, in this order: R2, and then for each segment R3,
// R4, R4b, R5, R6, R6b and R6c; then R10. R1 and R8 belong to the third
// layer, and R7 and R9 to the whole tree (CheckTree). The first violation
// is returned.
func CheckPath(path string) error {
	segs := strings.Split(path, "/")
	if len(segs) > MaxSegments {
		return fail("R2", "%d segments, more than %d", len(segs), MaxSegments)
	}
	for i, s := range segs {
		if s == "" {
			return fail("R2", "segment %d is empty", i+1)
		}
	}
	for i, s := range segs {
		if err := checkSegment(s); err != nil {
			e := err.(*Error)
			e.Detail = fmt.Sprintf("segment %d: %s", i+1, e.Detail)
			return e
		}
	}
	if strings.HasPrefix(Key(segs[0]), reservedDirPrefix) {
		return fail("R10", "the first segment starts with %q", reservedDirPrefix)
	}
	return nil
}

// checkSegment applies R3, R4, R4b, R5, R6, R6b and R6c to a segment.
func checkSegment(s string) error {
	if err := checkR3(s); err != nil {
		return err
	}
	for _, r := range s {
		if err := checkR4(r); err != nil {
			return err
		}
	}
	if err := checkPlacement("R4b", s); err != nil {
		return err
	}
	if err := checkR5(s); err != nil {
		return err
	}
	if err := checkR6(s); err != nil {
		return err
	}
	if err := checkR6b(s); err != nil {
		return err
	}
	return checkR6c(s)
}

func checkR3(s string) error {
	if len(s) > MaxSegmentLen {
		return fail("R3", "%d bytes, more than %d", len(s), MaxSegmentLen)
	}
	switch s {
	case ".":
		return fail("R3", "the segment is a dot")
	case "..":
		return fail("R3", "the segment is two dots")
	}
	switch stripWhitelist(s) {
	case "":
		return fail("R3", "the segment is empty without ZWNJ, ZWJ, VS15 and VS16")
	case ".":
		return fail("R3", "the segment is a dot without ZWNJ, ZWJ, VS15 and VS16")
	case "..":
		return fail("R3", "the segment is two dots without ZWNJ, ZWJ, VS15 and VS16")
	}
	if n := utf16Len(NFD(s)); n > MaxSegmentUTF16 {
		return fail("R3", "its NFD is %d UTF-16 code units, more than %d", n, MaxSegmentUTF16)
	}
	return nil
}

func utf16Len(s string) int {
	n := 0
	for _, r := range s {
		if r >= 0x10000 {
			n += 2
		} else {
			n++
		}
	}
	return n
}

// checkR4 rejects the code points of R4.
func checkR4(r rune) error {
	switch {
	case r <= 0x1F || (r >= 0x7F && r <= 0x9F):
		return fail("R4", "control %s", codePoint(r))
	case strings.ContainsRune(forbiddenASCII, r):
		return fail("R4", "character %s", codePoint(r))
	case r == 0x2028 || r == 0x2029:
		return fail("R4", "separator %s", codePoint(r))
	case DefaultIgnorable(r) && !whitelisted(r):
		return fail("R4", "invisible %s", codePoint(r))
	case r >= 0xF000 && r <= 0xF0FF:
		return fail("R4", "private use %s", codePoint(r))
	case !Assigned(r):
		return fail("R4", "unassigned %s", codePoint(r))
	}
	return nil
}

// checkPlacement applies R4b to s, a segment or a line of a text: VS15 and
// VS16 only right after a character that forms an emoji variation sequence
// with them, and ZWJ and ZWNJ never first, last or right after another ZWJ
// or ZWNJ.
func checkPlacement(rule, s string) error {
	prev, i := rune(-1), 0
	for _, r := range s {
		switch r {
		case vs15, vs16:
			if prev < 0 || !allowsSelector(prev, r) {
				return fail(rule, "%s is not part of an emoji variation sequence", codePoint(r))
			}
		case zwj, zwnj:
			switch {
			case prev < 0:
				return fail(rule, "%s at the start", codePoint(r))
			case prev == zwj || prev == zwnj:
				return fail(rule, "%s right after %s", codePoint(r), codePoint(prev))
			case i+utf8.RuneLen(r) == len(s):
				return fail(rule, "%s at the end", codePoint(r))
			}
		}
		prev = r
		i += utf8.RuneLen(r)
	}
	return nil
}

func checkR5(s string) error {
	switch {
	case strings.HasPrefix(s, " "):
		return fail("R5", "the segment starts with U+0020")
	case strings.HasSuffix(s, " "):
		return fail("R5", "the segment ends with U+0020")
	case strings.HasSuffix(s, "."):
		return fail("R5", "the segment ends with '.'")
	}
	return nil
}

// reserved are the device names of R6, in upper case.
var reserved = map[string]bool{
	"CON": true, "PRN": true, "AUX": true, "NUL": true, "CONIN$": true, "CONOUT$": true,
	"COM¹": true, "COM²": true, "COM³": true, "LPT¹": true, "LPT²": true, "LPT³": true,
}

func init() {
	for d := '0'; d <= '9'; d++ {
		reserved["COM"+string(d)] = true
		reserved["LPT"+string(d)] = true
	}
}

func checkR6(s string) error {
	base, _, _ := strings.Cut(s, ".")
	base = strings.TrimRight(base, " ")
	if name := asciiUpper(base); reserved[name] {
		return fail("R6", "%s is a reserved device name", name)
	}
	return nil
}

// asciiUpper upper-cases ASCII letters only.
func asciiUpper(s string) string {
	return strings.Map(func(r rune) rune {
		if r >= 'a' && r <= 'z' {
			return r - 'a' + 'A'
		}
		return r
	}, s)
}

// shortNameBase is the base of an 8.3 alias in R6b: it ends in '~' and 1 to
// 6 ASCII digits. It has no lookahead, which the regexp package lacks.
var shortNameBase = regexp.MustCompile(`^[^.]*~[0-9]{1,6}$`)

func checkR6b(s string) error {
	if strings.Count(s, ".") > 1 {
		return nil
	}
	base, ext, _ := strings.Cut(s, ".")
	if n := utf8.RuneCountInString(base); n < 1 || n > 8 {
		return nil
	}
	if utf8.RuneCountInString(ext) > 3 || !shortNameBase.MatchString(base) {
		return nil
	}
	return fail("R6b", "the segment has the form of an 8.3 alias")
}

// checkR6c projects the segment with each best-fit table: every non-ASCII
// code point the table maps to an ASCII byte is replaced by that byte. A
// projection must not hold '/', '\', ':' or U+0000, and must pass R3, R5, R6
// and R6b.
func checkR6c(s string) error {
	for i := range bestFitTables {
		t := &bestFitTables[i]
		p, changed := project(s, t)
		if !changed {
			continue
		}
		if j := strings.IndexAny(p, "/\\:\x00"); j >= 0 {
			return fail("R6c", "code page %s maps the segment to one with %s", t.name, codePoint(rune(p[j])))
		}
		for _, check := range []func(string) error{checkR3, checkR5, checkR6, checkR6b} {
			if err := check(p); err != nil {
				e := err.(*Error)
				return fail("R6c", "code page %s maps the segment to one that breaks %s: %s", t.name, e.Rule, e.Detail)
			}
		}
	}
	return nil
}

func project(s string, t *bestFitTable) (string, bool) {
	var b strings.Builder
	changed := false
	for _, r := range s {
		if r >= 0x80 {
			if c, ok := t.lookup(r); ok {
				b.WriteByte(c)
				changed = true
				continue
			}
		}
		b.WriteRune(r)
	}
	return b.String(), changed
}

// CheckTree applies R7 and then R9 to the paths of a head, which have passed
// CheckPath: no two siblings with the same key, no path that is both a file
// and a folder, comparing keys, and at most MaxImplicitDirs folders. Paths
// are numbered from 1 in the messages.
func CheckTree(paths []string) error {
	type node struct {
		name string
		dir  bool
		path int // the first path that reached it
	}
	// children maps the key of a folder, "" for the root, to its children by
	// key; the key of a folder joins the keys of its segments with '/'.
	children := map[string]map[string]node{}
	for n, path := range paths {
		segs := strings.Split(path, "/")
		parent := ""
		for i, s := range segs {
			k := Key(s)
			isDir := i < len(segs)-1
			kids := children[parent]
			if kids == nil {
				kids = map[string]node{}
				children[parent] = kids
			}
			switch old, ok := kids[k]; {
			case !ok:
				kids[k] = node{name: s, dir: isDir, path: n + 1}
			case old.name != s:
				return &Error{Rule: "R7", Detail: fmt.Sprintf("path %d collides with path %d in segment %d", n+1, old.path, i+1), Paths: [2]int{n + 1, old.path}}
			case old.dir != isDir || !isDir:
				return &Error{Rule: "R7", Detail: fmt.Sprintf("path %d makes a file of path %d a folder, or the reverse, in segment %d", n+1, old.path, i+1), Paths: [2]int{n + 1, old.path}}
			}
			if parent == "" {
				parent = k
			} else {
				parent += "/" + k
			}
		}
	}
	dirs := map[string]bool{}
	for _, path := range paths {
		for i := 0; ; {
			j := strings.IndexByte(path[i:], '/')
			if j < 0 {
				break
			}
			i += j
			dirs[path[:i]] = true
			i++
		}
	}
	if len(dirs) > MaxImplicitDirs {
		return fail("R9", "%d folders, more than %d", len(dirs), MaxImplicitDirs)
	}
	return nil
}

// CheckComment checks the comment of a head (spec §29.6).
func CheckComment(s string) error {
	return checkText(s, true)
}

// CheckAuthor checks the declared author of a head (spec §29.6).
func CheckAuthor(s string) error {
	if err := checkText(s, false); err != nil {
		return err
	}
	if strings.HasPrefix(s, " ") || strings.HasSuffix(s, " ") {
		return fail("text", "the declared author starts or ends with U+0020")
	}
	return nil
}

// checkText applies the rules of spec §29.6, code point by code point, and
// then R4b to each line.
func checkText(s string, comment bool) error {
	for _, r := range s {
		switch {
		case r == '\t' || r == '\n':
			if !comment {
				return fail("text", "control %s in the declared author", codePoint(r))
			}
		case r <= 0x1F || (r >= 0x7F && r <= 0x9F):
			return fail("text", "control %s", codePoint(r))
		case (r >= 0x202A && r <= 0x202E) || (r >= 0x2066 && r <= 0x2069) || r == 0x061C || r == 0x200E || r == 0x200F:
			return fail("text", "bidirectional control %s", codePoint(r))
		case r == 0x2028 || r == 0x2029:
			return fail("text", "separator %s", codePoint(r))
		case r == 0xFEFF:
			return fail("text", "byte order mark %s", codePoint(r))
		case isNoncharacter(r):
			return fail("text", "noncharacter %s", codePoint(r))
		case DefaultIgnorable(r) && !whitelisted(r):
			return fail("text", "invisible %s", codePoint(r))
		}
	}
	for i, line := range strings.Split(s, "\n") {
		if err := checkPlacement("text", line); err != nil {
			e := err.(*Error)
			e.Detail = fmt.Sprintf("line %d: %s", i+1, e.Detail)
			return e
		}
	}
	return nil
}

// isNoncharacter reports the 66 noncharacters: U+FDD0 to U+FDEF, and the last
// two code points of each plane.
func isNoncharacter(r rune) bool {
	return (r >= 0xFDD0 && r <= 0xFDEF) || r&0xFFFE == 0xFFFE
}