format_state.go

package do

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

var (
	_ = fmt.Printf
)

// ===== Code copied from github.com/golang/tools/go/analysis/passes/printf/printf.go
// ===== and simplified for checking string if is a format.

// formatState holds the parsed representation of a printf directive such as "%3.*[4]d".
// It is constructed by parsePrintfVerb.
type formatState struct {
	verb     rune   // the format verb: 'd' for "%d"
	format   string // the full format directive from % through verb, "%.3d".
	name     string // Printf, Sprintf etc.
	flags    []byte // the list of # + etc.
	argNums  []int  // the successive argument numbers that are consumed, adjusted to refer to actual arg in call
	firstArg int    // Index of first argument after the format in the Printf call.
	// Used only during parse.
	argNum       int  // Which argument we're expecting to format now.
	hasIndex     bool // Whether the argument is indexed.
	indexPending bool // Whether we have an indexed argument that has not resolved.
	nbytes       int  // number of bytes of the format string consumed.
}

// checkPrintf checks a call to a formatted print routine such as Printf.
func checkPrintf(format string) (ok bool, argNum int) {
	idx := 0
	firstArg := idx // Arguments are immediately after format string.
	if !strings.Contains(format, "%") {
		return
	}
	// fmt.Printf("= format: %s\n", format)
	// Hard part: check formats against args.
	argNum = firstArg
	maxArgNum := firstArg
	anyIndex := false
	for i, w := 0, 0; i < len(format); i += w {
		w = 1
		if format[i] != '%' {
			continue
		}
		state := parsePrintfVerb(format[i:], firstArg, argNum)
		if state == nil {
			return
		}
		// fmt.Printf("=== verb: %s\n", string(state.verb))
		w = len(state.format)
		if !okPrintfArg(state) { // One error per format is enough.
			continue // change from return to continue to get left verb
		}
		if state.hasIndex {
			anyIndex = true
		}

		if len(state.argNums) > 0 {
			// Continue with the next sequential argument.
			argNum = state.argNums[len(state.argNums)-1] + 1
		}
		for _, n := range state.argNums {
			if n >= maxArgNum {
				maxArgNum = n + 1
			}
		}
	}

	if anyIndex {
		return
	}

	return true, argNum
}

// parseFlags accepts any printf flags.
func (s *formatState) parseFlags() {
	for s.nbytes < len(s.format) {
		switch c := s.format[s.nbytes]; c {
		case '#', '0', '+', '-', ' ':
			s.flags = append(s.flags, c)
			s.nbytes++
		default:
			return
		}
	}
}

// scanNum advances through a decimal number if present.
func (s *formatState) scanNum() {
	for ; s.nbytes < len(s.format); s.nbytes++ {
		c := s.format[s.nbytes]
		if c < '0' || '9' < c {
			return
		}
	}
}

// parseIndex scans an index expression. It returns false if there is a syntax error.
func (s *formatState) parseIndex() bool {
	if s.nbytes == len(s.format) || s.format[s.nbytes] != '[' {
		return true
	}
	// Argument index present.
	s.nbytes++ // skip '['
	start := s.nbytes
	s.scanNum()
	ok := true
	if s.nbytes == len(s.format) || s.nbytes == start || s.format[s.nbytes] != ']' {
		ok = false // syntax error is either missing "]" or invalid index.
		s.nbytes = strings.Index(s.format[start:], "]")
		if s.nbytes < 0 {
			// s.pass.ReportRangef(s.call, "%s format %s is missing closing ]", s.name, s.format)
			return false
		}
		s.nbytes = s.nbytes + start
	}
	_ = ok
	arg32, err := strconv.ParseInt(s.format[start:s.nbytes], 10, 32)
	_ = err

	s.nbytes++ // skip ']'
	arg := int(arg32)
	arg += s.firstArg - 1 // We want to zero-index the actual arguments.
	s.argNum = arg
	s.hasIndex = true
	s.indexPending = true
	return true
}

// parseNum scans a width or precision (or *). It returns false if there's a bad index expression.
func (s *formatState) parseNum() bool {
	if s.nbytes < len(s.format) && s.format[s.nbytes] == '*' {
		if s.indexPending { // Absorb it.
			s.indexPending = false
		}
		s.nbytes++
		s.argNums = append(s.argNums, s.argNum)
		s.argNum++
	} else {
		s.scanNum()
	}
	return true
}

// parsePrecision scans for a precision. It returns false if there's a bad index expression.
func (s *formatState) parsePrecision() bool {
	// If there's a period, there may be a precision.
	if s.nbytes < len(s.format) && s.format[s.nbytes] == '.' {
		s.flags = append(s.flags, '.') // Treat precision as a flag.
		s.nbytes++
		if !s.parseIndex() {
			return false
		}
		if !s.parseNum() {
			return false
		}
	}
	return true
}

// parsePrintfVerb looks the formatting directive that begins the format string
// and returns a formatState that encodes what the directive wants, without looking
// at the actual arguments present in the call. The result is nil if there is an error.
func parsePrintfVerb(format string, firstArg, argNum int) *formatState {
	state := &formatState{
		format:   format,
		name:     format,
		flags:    make([]byte, 0, 5),
		argNum:   argNum,
		argNums:  make([]int, 0, 1),
		nbytes:   1, // There's guaranteed to be a percent sign.
		firstArg: firstArg,
	}
	// There may be flags.
	state.parseFlags()
	// There may be an index.
	if !state.parseIndex() {
		return nil
	}
	// There may be a width.
	if !state.parseNum() {
		return nil
	}
	// There may be a precision.
	if !state.parsePrecision() {
		return nil
	}
	// Now a verb, possibly prefixed by an index (which we may already have).
	if !state.indexPending && !state.parseIndex() {
		return nil
	}
	if state.nbytes == len(state.format) {
		return nil
	}
	verb, w := utf8.DecodeRuneInString(state.format[state.nbytes:])
	state.verb = verb
	state.nbytes += w
	if verb != '%' {
		state.argNums = append(state.argNums, state.argNum)
	}
	state.format = state.format[:state.nbytes]
	return state
}

// printfArgType encodes the types of expressions a printf verb accepts. It is a bitmask.
type printfArgType int

const (
	argBool printfArgType = 1 << iota
	argInt
	argRune
	argString
	argFloat
	argComplex
	argPointer
	argError
	anyType printfArgType = ^0
)

type printVerb struct {
	verb  rune   // User may provide verb through Formatter; could be a rune.
	flags string // known flags are all ASCII
	typ   printfArgType
}

// Common flag sets for printf verbs.
const (
	noFlag       = ""
	numFlag      = " -+.0"
	sharpNumFlag = " -+.0#"
	allFlags     = " -+.0#"
)

// printVerbs identifies which flags are known to printf for each verb.
var printVerbs = []printVerb{
	// '-' is a width modifier, always valid.
	// '.' is a precision for float, max width for strings.
	// '+' is required sign for numbers, Go format for %v.
	// '#' is alternate format for several verbs.
	// ' ' is spacer for numbers
	{'%', noFlag, 0},
	{'b', sharpNumFlag, argInt | argFloat | argComplex | argPointer},
	{'c', "-", argRune | argInt},
	{'d', numFlag, argInt | argPointer},
	{'e', sharpNumFlag, argFloat | argComplex},
	{'E', sharpNumFlag, argFloat | argComplex},
	{'f', sharpNumFlag, argFloat | argComplex},
	{'F', sharpNumFlag, argFloat | argComplex},
	{'g', sharpNumFlag, argFloat | argComplex},
	{'G', sharpNumFlag, argFloat | argComplex},
	{'o', sharpNumFlag, argInt | argPointer},
	{'O', sharpNumFlag, argInt | argPointer},
	{'p', "-#", argPointer},
	{'q', " -+.0#", argRune | argInt | argString},
	{'s', " -+.0", argString},
	{'t', "-", argBool},
	{'T', "-", anyType},
	{'U', "-#", argRune | argInt},
	{'v', allFlags, anyType},
	{'w', allFlags, argError},
	{'x', sharpNumFlag, argRune | argInt | argString | argPointer | argFloat | argComplex},
	{'X', sharpNumFlag, argRune | argInt | argString | argPointer | argFloat | argComplex},
}

// okPrintfArg compares the formatState to the arguments actually present,
// reporting any discrepancies it can discern. If the final argument is ellipsissed,
// there's little it can do for that.
func okPrintfArg(state *formatState) (ok bool) {
	var v printVerb
	found := false
	// Linear scan is fast enough for a small list.
	for _, v = range printVerbs {
		if v.verb == state.verb {
			found = true
			break
		}
	}

	// Could current arg implement fmt.Formatter?
	// Skip check for the %w verb, which requires an error.
	formatter := false

	if !formatter {
		if !found {
			return false
		}
		for _, flag := range state.flags {
			if flag == '0' {
				continue
			}
			if !strings.ContainsRune(v.flags, rune(flag)) {
				return false
			}
		}
	}

	if state.verb == '%' || formatter {
		return true
	}

	return true
}