Source: js_printer.go in package github.com/evanw/esbuild/internal/js

package js_printer

import (
	"bytes"
	"fmt"
	"math"
	"strconv"
	"strings"
	"unicode/utf8"

	"github.com/evanw/esbuild/internal/ast"
	"github.com/evanw/esbuild/internal/compat"
	"github.com/evanw/esbuild/internal/config"
	"github.com/evanw/esbuild/internal/js_ast"
	"github.com/evanw/esbuild/internal/js_lexer"
	"github.com/evanw/esbuild/internal/logger"
	"github.com/evanw/esbuild/internal/renamer"
	"github.com/evanw/esbuild/internal/sourcemap"
)

var positiveInfinity = math.Inf(1)
var negativeInfinity = math.Inf(-1)

Coordinates in source maps are stored using relative offsets for size reasons. When joining together chunks of a source map that were emitted in parallel for different parts of a file, we need to fix up the first segment of each chunk to be relative to the end of the previous chunk.

This isn't stored in the source map. It's only used by the bundler to join source map chunks together correctly.

	GeneratedLine int

These are stored in the source map in VLQ format.

	GeneratedColumn int
	SourceIndex     int
	OriginalLine    int
	OriginalColumn  int
}

Source map chunks are computed in parallel for speed. Each chunk is relative to the zero state instead of being relative to the end state of the previous chunk, since it's impossible to know the end state of the previous chunk in a parallel computation. After all chunks are computed, they are joined together in a second pass. This rewrites the first mapping in each chunk to be relative to the end state of the previous chunk.

Handle line breaks in between this mapping and the previous one

	if startState.GeneratedLine != 0 {
		j.AddBytes(bytes.Repeat([]byte{';'}, startState.GeneratedLine))
		prevEndState.GeneratedColumn = 0
	}

Skip past any leading semicolons, which indicate line breaks

	semicolons := 0
	for sourceMap[semicolons] == ';' {
		semicolons++
	}
	if semicolons > 0 {
		j.AddBytes(sourceMap[:semicolons])
		sourceMap = sourceMap[semicolons:]
		prevEndState.GeneratedColumn = 0
		startState.GeneratedColumn = 0
	}

Strip off the first mapping from the buffer. The first mapping should be for the start of the original file (the printer always generates one for the start of the file).

	generatedColumn, i := sourcemap.DecodeVLQ(sourceMap, 0)
	sourceIndex, i := sourcemap.DecodeVLQ(sourceMap, i)
	originalLine, i := sourcemap.DecodeVLQ(sourceMap, i)
	originalColumn, i := sourcemap.DecodeVLQ(sourceMap, i)
	sourceMap = sourceMap[i:]

Rewrite the first mapping to be relative to the end state of the previous chunk. We now know what the end state is because we're in the second pass where all chunks have already been generated.

	startState.SourceIndex += sourceIndex
	startState.GeneratedColumn += generatedColumn
	startState.OriginalLine += originalLine
	startState.OriginalColumn += originalColumn
	j.AddBytes(appendMapping(nil, j.lastByte, prevEndState, startState))

Then append everything after that without modification.

	j.AddBytes(sourceMap)
}

Put commas in between mappings

	if lastByte != 0 && lastByte != ';' && lastByte != '"' {
		buffer = append(buffer, ',')
	}

Record the generated column (the line is recorded using ';' elsewhere)

	buffer = append(buffer, sourcemap.EncodeVLQ(currentState.GeneratedColumn-prevState.GeneratedColumn)...)
	prevState.GeneratedColumn = currentState.GeneratedColumn

Record the generated source

	buffer = append(buffer, sourcemap.EncodeVLQ(currentState.SourceIndex-prevState.SourceIndex)...)
	prevState.SourceIndex = currentState.SourceIndex

Record the original line

	buffer = append(buffer, sourcemap.EncodeVLQ(currentState.OriginalLine-prevState.OriginalLine)...)
	prevState.OriginalLine = currentState.OriginalLine

Record the original column

	buffer = append(buffer, sourcemap.EncodeVLQ(currentState.OriginalColumn-prevState.OriginalColumn)...)
	prevState.OriginalColumn = currentState.OriginalColumn

	return buffer
}

This provides an efficient way to join lots of big string and byte slices together. It avoids the cost of repeatedly reallocating as the buffer grows by measuring exactly how big the buffer should be and then allocating once. This is a measurable speedup.

type Joiner struct {
	lastByte byte
	strings  []joinerString
	bytes    []joinerBytes
	length   uint32
}

type joinerString struct {
	data   string
	offset uint32
}

type joinerBytes struct {
	data   []byte
	offset uint32
}

func (j *Joiner) AddString(data string) {
	if len(data) > 0 {
		j.lastByte = data[len(data)-1]
	}
	j.strings = append(j.strings, joinerString{data, j.length})
	j.length += uint32(len(data))
}

func (j *Joiner) AddBytes(data []byte) {
	if len(data) > 0 {
		j.lastByte = data[len(data)-1]
	}
	j.bytes = append(j.bytes, joinerBytes{data, j.length})
	j.length += uint32(len(data))
}

func (j *Joiner) LastByte() byte {
	return j.lastByte
}

func (j *Joiner) Length() uint32 {
	return j.length
}

func (j *Joiner) Done() []byte {
	buffer := make([]byte, j.length)
	for _, item := range j.strings {
		copy(buffer[item.offset:], item.data)
	}
	for _, item := range j.bytes {
		copy(buffer[item.offset:], item.data)
	}
	return buffer
}

const hexChars = "0123456789ABCDEF"
const firstASCII = 0x20
const lastASCII = 0x7E
const firstHighSurrogate = 0xD800
const lastHighSurrogate = 0xDBFF
const firstLowSurrogate = 0xDC00
const lastLowSurrogate = 0xDFFF

func canPrintWithoutEscape(c rune, asciiOnly bool) bool {
	if c <= lastASCII {
		return c >= firstASCII && c != '\\' && c != '"'
	} else {
		return !asciiOnly && c != '\uFEFF' && (c < firstHighSurrogate || c > lastLowSurrogate)
	}
}

Estimate the required length

	lenEstimate := 2
	for _, c := range text {
		if canPrintWithoutEscape(c, asciiOnly) {
			lenEstimate += utf8.RuneLen(c)
		} else {
			switch c {
			case '\b', '\f', '\n', '\r', '\t', '\\', '"':
				lenEstimate += 2
			default:
				if c <= 0xFFFF {
					lenEstimate += 6
				} else {
					lenEstimate += 12
				}
			}
		}
	}

Preallocate the array

	bytes := make([]byte, 0, lenEstimate)
	i := 0
	n := len(text)
	bytes = append(bytes, '"')

	for i < n {
		c, width := js_lexer.DecodeWTF8Rune(text[i:])

Fast path: a run of characters that don't need escaping

		if canPrintWithoutEscape(c, asciiOnly) {
			start := i
			i += width
			for i < n {
				c, width = js_lexer.DecodeWTF8Rune(text[i:])
				if !canPrintWithoutEscape(c, asciiOnly) {
					break
				}
				i += width
			}
			bytes = append(bytes, text[start:i]...)
			continue
		}

		switch c {
		case '\b':
			bytes = append(bytes, "\\b"...)
			i++

		case '\f':
			bytes = append(bytes, "\\f"...)
			i++

		case '\n':
			bytes = append(bytes, "\\n"...)
			i++

		case '\r':
			bytes = append(bytes, "\\r"...)
			i++

		case '\t':
			bytes = append(bytes, "\\t"...)
			i++

		case '\\':
			bytes = append(bytes, "\\\\"...)
			i++

		case '"':
			bytes = append(bytes, "\\\""...)
			i++

		default:
			i += width
			if c <= 0xFFFF {
				bytes = append(
					bytes,
					'\\', 'u', hexChars[c>>12], hexChars[(c>>8)&15], hexChars[(c>>4)&15], hexChars[c&15],
				)
			} else {
				c -= 0x10000
				lo := firstHighSurrogate + ((c >> 10) & 0x3FF)
				hi := firstLowSurrogate + (c & 0x3FF)
				bytes = append(
					bytes,
					'\\', 'u', hexChars[lo>>12], hexChars[(lo>>8)&15], hexChars[(lo>>4)&15], hexChars[lo&15],
					'\\', 'u', hexChars[hi>>12], hexChars[(hi>>8)&15], hexChars[(hi>>4)&15], hexChars[hi&15],
				)
			}
		}
	}

	return append(bytes, '"')
}

func QuoteIdentifier(js []byte, name string, unsupportedFeatures compat.JSFeature) []byte {
	isASCII := false
	asciiStart := 0
	for i, c := range name {

Fast path: a run of ASCII characters

			if !isASCII {
				isASCII = true
				asciiStart = i
			}

Slow path: escape non-ACSII characters

			if isASCII {
				js = append(js, name[asciiStart:i]...)
				isASCII = false
			}
			if c <= 0xFFFF {
				js = append(js, '\\', 'u', hexChars[c>>12], hexChars[(c>>8)&15], hexChars[(c>>4)&15], hexChars[c&15])
			} else if !unsupportedFeatures.Has(compat.UnicodeEscapes) {
				js = append(js, fmt.Sprintf("\\u{%X}", c)...)
			} else {
				panic("Internal error: Cannot encode identifier: Unicode escapes are unsupported")
			}
		}
	}

Print one final run of ASCII characters

		js = append(js, name[asciiStart:]...)
	}
	return js
}

func (p *printer) printQuotedUTF16(text []uint16, quote rune) {
	temp := make([]byte, utf8.UTFMax)
	js := p.js
	i := 0
	n := len(text)

	for i < n {
		c := text[i]
		i++

Special-case the null character since it may mess with code written in C that treats null characters as the end of the string.

We don't want "\x001" to be written as "\01"

			if i < n && text[i] >= '0' && text[i] <= '9' {
				js = append(js, "\\x00"...)
			} else {
				js = append(js, "\\0"...)
			}

Special-case the bell character since it may cause dumping this file to the terminal to make a sound, which is undesirable. Note that we can't use an octal literal to print this shorter since octal literals are not allowed in strict mode (or in template strings).

		case '\x07':
			js = append(js, "\\x07"...)

		case '\b':
			js = append(js, "\\b"...)

		case '\f':
			js = append(js, "\\f"...)

		case '\n':
			if quote == '`' {
				js = append(js, '\n')
			} else {
				js = append(js, "\\n"...)
			}

		case '\r':
			js = append(js, "\\r"...)

		case '\v':
			js = append(js, "\\v"...)

		case '\\':
			js = append(js, "\\\\"...)

		case '\'':
			if quote == '\'' {
				js = append(js, '\\')
			}
			js = append(js, '\'')

		case '"':
			if quote == '"' {
				js = append(js, '\\')
			}
			js = append(js, '"')

		case '`':
			if quote == '`' {
				js = append(js, '\\')
			}
			js = append(js, '`')

		case '$':
			if quote == '`' && i < n && text[i] == '{' {
				js = append(js, '\\')
			}
			js = append(js, '$')

		case '\u2028':
			js = append(js, "\\u2028"...)

		case '\u2029':
			js = append(js, "\\u2029"...)

		case '\uFEFF':
			js = append(js, "\\uFEFF"...)

		default:

Common case: just append a single byte

			case c <= lastASCII:
				js = append(js, byte(c))

Is this a high surrogate?

Is there a next character?

				if i < n {
					c2 := text[i]

Is it a low surrogate?

					if c2 >= firstLowSurrogate && c2 <= lastLowSurrogate {
						r := (rune(c) << 10) + rune(c2) + (0x10000 - (firstHighSurrogate << 10) - firstLowSurrogate)
						i++

Escape this character if UTF-8 isn't allowed

						if p.options.ASCIIOnly {
							if !p.options.UnsupportedFeatures.Has(compat.UnicodeEscapes) {
								js = append(js, fmt.Sprintf("\\u{%X}", r)...)
							} else {
								js = append(js,
									'\\', 'u', hexChars[c>>12], hexChars[(c>>8)&15], hexChars[(c>>4)&15], hexChars[c&15],
									'\\', 'u', hexChars[c2>>12], hexChars[(c2>>8)&15], hexChars[(c2>>4)&15], hexChars[c2&15],
								)
							}
							continue
						}

Otherwise, encode to UTF-8

						width := utf8.EncodeRune(temp, r)
						js = append(js, temp[:width]...)
						continue
					}
				}

Write an unpaired high surrogate

				js = append(js, '\\', 'u', hexChars[c>>12], hexChars[(c>>8)&15], hexChars[(c>>4)&15], hexChars[c&15])

Is this an unpaired low surrogate or four-digit hex escape?

			case (c >= firstLowSurrogate && c <= lastLowSurrogate) || (p.options.ASCIIOnly && c > 0xFF):
				js = append(js, '\\', 'u', hexChars[c>>12], hexChars[(c>>8)&15], hexChars[(c>>4)&15], hexChars[c&15])

Can this be a two-digit hex escape?

			case p.options.ASCIIOnly:
				js = append(js, '\\', 'x', hexChars[c>>4], hexChars[c&15])

Otherwise, just encode to UTF-8

			default:
				width := utf8.EncodeRune(temp, rune(c))
				js = append(js, temp[:width]...)
			}
		}
	}

	p.js = js
}

type printer struct {
	symbols            js_ast.SymbolMap
	renamer            renamer.Renamer
	importRecords      []ast.ImportRecord
	options            Options
	extractedComments  map[string]bool
	needsSemicolon     bool
	js                 []byte
	stmtStart          int
	exportDefaultStart int
	arrowExprStart     int
	forOfInitStart     int
	prevOp             js_ast.OpCode
	prevOpEnd          int
	prevNumEnd         int
	prevRegExpEnd      int
	callTarget         js_ast.E
	intToBytesBuffer   [64]byte

For source maps

	sourceMap           []byte
	prevLoc             logger.Loc
	prevState           SourceMapState
	lastGeneratedUpdate int
	generatedColumn     int
	hasPrevState        bool
	lineOffsetTables    []LineOffsetTable

This is a workaround for a bug in the popular "source-map" library: https://github.com/mozilla/source-map/issues/261. The library will sometimes return null when querying a source map unless every line starts with a mapping at column zero. The workaround is to replicate the previous mapping if a line ends up not starting with a mapping. This is done lazily because we want to avoid replicating the previous mapping if we don't need to.

	lineStartsWithMapping     bool
	coverLinesWithoutMappings bool
}

type LineOffsetTable struct {
	byteOffsetToStartOfLine int32

The source map specification is very loose and does not specify what column numbers actually mean. The popular "source-map" library from Mozilla appears to interpret them as counts of UTF-16 code units, so we generate those too for compatibility. We keep mapping tables around to accelerate conversion from byte offsets to UTF-16 code unit counts. However, this mapping takes up a lot of memory and generates a lot of garbage. Since most JavaScript is ASCII and the mapping for ASCII is 1:1, we avoid creating a table for ASCII-only lines as an optimization.

	byteOffsetToFirstNonASCII int32
	columnsForNonASCII        []int32
}

func (p *printer) print(text string) {
	p.js = append(p.js, text...)
}

This is the same as "print(string(bytes))" without any unnecessary temporary allocations

func (p *printer) printBytes(bytes []byte) {
	p.js = append(p.js, bytes...)
}

func (p *printer) printQuotedUTF8(text string, allowBacktick bool) {
	value := js_lexer.StringToUTF16(text)
	c := p.bestQuoteCharForString(value, allowBacktick)
	p.print(c)
	p.printQuotedUTF16(value, rune(c[0]))
	p.print(c)
}

func (p *printer) addSourceMapping(loc logger.Loc) {
	if !p.options.AddSourceMappings || loc == p.prevLoc {
		return
	}
	p.prevLoc = loc

Binary search to find the line

	lineOffsetTables := p.lineOffsetTables
	count := len(lineOffsetTables)
	originalLine := 0
	for count > 0 {
		step := count / 2
		i := originalLine + step
		if lineOffsetTables[i].byteOffsetToStartOfLine <= loc.Start {
			originalLine = i + 1
			count = count - step - 1
		} else {
			count = step
		}
	}
	originalLine--

Use the line to compute the column

	line := &lineOffsetTables[originalLine]
	originalColumn := int(loc.Start - line.byteOffsetToStartOfLine)
	if line.columnsForNonASCII != nil && originalColumn >= int(line.byteOffsetToFirstNonASCII) {
		originalColumn = int(line.columnsForNonASCII[originalColumn-int(line.byteOffsetToFirstNonASCII)])
	}

	p.updateGeneratedLineAndColumn()

If this line doesn't start with a mapping and we're about to add a mapping that's not at the start, insert a mapping first so the line starts with one.

	if p.coverLinesWithoutMappings && !p.lineStartsWithMapping && p.generatedColumn > 0 && p.hasPrevState {
		p.appendMappingWithoutRemapping(SourceMapState{
			GeneratedLine:   p.prevState.GeneratedLine,
			GeneratedColumn: 0,
			SourceIndex:     p.prevState.SourceIndex,
			OriginalLine:    p.prevState.OriginalLine,
			OriginalColumn:  p.prevState.OriginalColumn,
		})
	}

	p.appendMapping(SourceMapState{
		GeneratedLine:   p.prevState.GeneratedLine,
		GeneratedColumn: p.generatedColumn,
		OriginalLine:    originalLine,
		OriginalColumn:  originalColumn,
	})

This line now has a mapping on it, so don't insert another one

	p.lineStartsWithMapping = true
}

Scan over the printed text since the last source mapping and update the generated line and column numbers

func (p *printer) updateGeneratedLineAndColumn() {
	for i, c := range string(p.js[p.lastGeneratedUpdate:]) {
		switch c {

Handle Windows-specific "\r\n" newlines

			if c == '\r' {
				newlineCheck := p.lastGeneratedUpdate + i + 1
				if newlineCheck < len(p.js) && p.js[newlineCheck] == '\n' {
					continue
				}
			}

If we're about to move to the next line and the previous line didn't have any mappings, add a mapping at the start of the previous line.

			if p.coverLinesWithoutMappings && !p.lineStartsWithMapping && p.hasPrevState {
				p.appendMappingWithoutRemapping(SourceMapState{
					GeneratedLine:   p.prevState.GeneratedLine,
					GeneratedColumn: 0,
					SourceIndex:     p.prevState.SourceIndex,
					OriginalLine:    p.prevState.OriginalLine,
					OriginalColumn:  p.prevState.OriginalColumn,
				})
			}

			p.prevState.GeneratedLine++
			p.prevState.GeneratedColumn = 0
			p.generatedColumn = 0
			p.sourceMap = append(p.sourceMap, ';')

This new line doesn't have a mapping yet

			p.lineStartsWithMapping = false

Mozilla's "source-map" library counts columns using UTF-16 code units

			if c <= 0xFFFF {
				p.generatedColumn++
			} else {
				p.generatedColumn += 2
			}
		}
	}

	p.lastGeneratedUpdate = len(p.js)
}

func GenerateLineOffsetTables(contents string, approximateLineCount int32) []LineOffsetTable {
	var columnsForNonASCII []int32
	byteOffsetToFirstNonASCII := int32(0)
	lineByteOffset := 0
	columnByteOffset := 0
	column := int32(0)

Preallocate the top-level table using the approximate line count from the lexer

	lineOffsetTables := make([]LineOffsetTable, 0, approximateLineCount)

Mark the start of the next line

		if column == 0 {
			lineByteOffset = i
		}

Start the mapping if this character is non-ASCII

		if c > 0x7F && columnsForNonASCII == nil {
			columnByteOffset = i - lineByteOffset
			byteOffsetToFirstNonASCII = int32(columnByteOffset)
			columnsForNonASCII = []int32{}
		}

Update the per-byte column offsets

		if columnsForNonASCII != nil {
			for lineBytesSoFar := i - lineByteOffset; columnByteOffset <= lineBytesSoFar; columnByteOffset++ {
				columnsForNonASCII = append(columnsForNonASCII, column)
			}
		}

		switch c {

Handle Windows-specific "\r\n" newlines

			if c == '\r' {
				if i+1 < len(contents) && contents[i+1] == '\n' {
					column++
					continue
				}
			}

			lineOffsetTables = append(lineOffsetTables, LineOffsetTable{
				byteOffsetToStartOfLine:   int32(lineByteOffset),
				byteOffsetToFirstNonASCII: byteOffsetToFirstNonASCII,
				columnsForNonASCII:        columnsForNonASCII,
			})
			columnByteOffset = 0
			byteOffsetToFirstNonASCII = 0
			columnsForNonASCII = nil
			column = 0

Mozilla's "source-map" library counts columns using UTF-16 code units

			if c <= 0xFFFF {
				column++
			} else {
				column += 2
			}
		}
	}

Mark the start of the next line

	if column == 0 {
		lineByteOffset = len(contents)
	}

Do one last update for the column at the end of the file

	if columnsForNonASCII != nil {
		for lineBytesSoFar := len(contents) - lineByteOffset; columnByteOffset <= lineBytesSoFar; columnByteOffset++ {
			columnsForNonASCII = append(columnsForNonASCII, column)
		}
	}

	lineOffsetTables = append(lineOffsetTables, LineOffsetTable{
		byteOffsetToStartOfLine:   int32(lineByteOffset),
		byteOffsetToFirstNonASCII: byteOffsetToFirstNonASCII,
		columnsForNonASCII:        columnsForNonASCII,
	})
	return lineOffsetTables
}

If the input file had a source map, map all the way back to the original

	if p.options.InputSourceMap != nil {
		mapping := p.options.InputSourceMap.Find(
			int32(currentState.OriginalLine),
			int32(currentState.OriginalColumn))

Some locations won't have a mapping

		if mapping == nil {
			return
		}

		currentState.SourceIndex = int(mapping.SourceIndex)
		currentState.OriginalLine = int(mapping.OriginalLine)
		currentState.OriginalColumn = int(mapping.OriginalColumn)
	}

	p.appendMappingWithoutRemapping(currentState)
}

func (p *printer) appendMappingWithoutRemapping(currentState SourceMapState) {
	var lastByte byte
	if len(p.sourceMap) != 0 {
		lastByte = p.sourceMap[len(p.sourceMap)-1]
	}

	p.sourceMap = appendMapping(p.sourceMap, lastByte, p.prevState, currentState)
	p.prevState = currentState
	p.hasPrevState = true
}

func (p *printer) printIndent() {
	if !p.options.RemoveWhitespace {
		for i := 0; i < p.options.Indent; i++ {
			p.print("  ")
		}
	}
}

func (p *printer) printSymbol(ref js_ast.Ref) {
	p.printSpaceBeforeIdentifier()
	p.printIdentifier(p.renamer.NameForSymbol(ref))
}

func CanQuoteIdentifier(name string, unsupportedJSFeatures compat.JSFeature, asciiOnly bool) bool {
	return js_lexer.IsIdentifier(name) && (!asciiOnly ||
		!unsupportedJSFeatures.Has(compat.UnicodeEscapes) ||
		!js_lexer.ContainsNonBMPCodePoint(name))
}

func (p *printer) canPrintIdentifier(name string) bool {
	return js_lexer.IsIdentifier(name) && (!p.options.ASCIIOnly ||
		!p.options.UnsupportedFeatures.Has(compat.UnicodeEscapes) ||
		!js_lexer.ContainsNonBMPCodePoint(name))
}

func (p *printer) canPrintIdentifierUTF16(name []uint16) bool {
	return js_lexer.IsIdentifierUTF16(name) && (!p.options.ASCIIOnly ||
		!p.options.UnsupportedFeatures.Has(compat.UnicodeEscapes) ||
		!js_lexer.ContainsNonBMPCodePointUTF16(name))
}

func (p *printer) printIdentifier(name string) {
	if p.options.ASCIIOnly {
		p.js = QuoteIdentifier(p.js, name, p.options.UnsupportedFeatures)
	} else {
		p.print(name)
	}
}

This is the same as "printIdentifier(StringToUTF16(bytes))" without any unnecessary temporary allocations

func (p *printer) printIdentifierUTF16(name []uint16) {
	temp := make([]byte, utf8.UTFMax)
	n := len(name)

	for i := 0; i < n; i++ {
		c := rune(name[i])

		if c >= firstHighSurrogate && c <= lastHighSurrogate && i+1 < n {
			if c2 := rune(name[i+1]); c2 >= firstLowSurrogate && c2 <= lastLowSurrogate {
				c = (c << 10) + c2 + (0x10000 - (firstHighSurrogate << 10) - firstLowSurrogate)
				i++
			}
		}

		if p.options.ASCIIOnly && c > lastASCII {
			if c <= 0xFFFF {
				p.js = append(p.js, '\\', 'u', hexChars[c>>12], hexChars[(c>>8)&15], hexChars[(c>>4)&15], hexChars[c&15])
			} else if !p.options.UnsupportedFeatures.Has(compat.UnicodeEscapes) {
				p.js = append(p.js, fmt.Sprintf("\\u{%X}", c)...)
			} else {
				panic("Internal error: Cannot encode identifier: Unicode escapes are unsupported")
			}
			continue
		}

		width := utf8.EncodeRune(temp, c)
		p.js = append(p.js, temp[:width]...)
	}
}

func (p *printer) printBinding(binding js_ast.Binding) {
	p.addSourceMapping(binding.Loc)

	switch b := binding.Data.(type) {
	case *js_ast.BMissing:

	case *js_ast.BIdentifier:
		p.printSymbol(b.Ref)

	case *js_ast.BArray:
		p.print("[")
		if len(b.Items) > 0 {
			if !b.IsSingleLine {
				p.options.Indent++
			}

			for i, item := range b.Items {
				if i != 0 {
					p.print(",")
					if b.IsSingleLine {
						p.printSpace()
					}
				}
				if !b.IsSingleLine {
					p.printNewline()
					p.printIndent()
				}
				if b.HasSpread && i+1 == len(b.Items) {
					p.print("...")
				}
				p.printBinding(item.Binding)

				if item.DefaultValue != nil {
					p.printSpace()
					p.print("=")
					p.printSpace()
					p.printExpr(*item.DefaultValue, js_ast.LComma, 0)
				}

Make sure there's a comma after trailing missing items

				if _, ok := item.Binding.Data.(*js_ast.BMissing); ok && i == len(b.Items)-1 {
					p.print(",")
				}
			}

			if !b.IsSingleLine {
				p.options.Indent--
				p.printNewline()
				p.printIndent()
			}
		}
		p.print("]")

	case *js_ast.BObject:
		p.print("{")
		if len(b.Properties) > 0 {
			if !b.IsSingleLine {
				p.options.Indent++
			}

			for i, property := range b.Properties {
				if i != 0 {
					p.print(",")
					if b.IsSingleLine {
						p.printSpace()
					}
				}
				if !b.IsSingleLine {
					p.printNewline()
					p.printIndent()
				}

				if property.IsSpread {
					p.print("...")
				} else {
					if property.IsComputed {
						p.print("[")
						p.printExpr(property.Key, js_ast.LComma, 0)
						p.print("]:")
						p.printSpace()
						p.printBinding(property.Value)

						if property.DefaultValue != nil {
							p.printSpace()
							p.print("=")
							p.printSpace()
							p.printExpr(*property.DefaultValue, js_ast.LComma, 0)
						}
						continue
					}

					if str, ok := property.Key.Data.(*js_ast.EString); ok {
						if p.canPrintIdentifierUTF16(str.Value) {
							p.addSourceMapping(property.Key.Loc)
							p.printSpaceBeforeIdentifier()
							p.printIdentifierUTF16(str.Value)

Use a shorthand property if the names are the same

							if id, ok := property.Value.Data.(*js_ast.BIdentifier); ok && js_lexer.UTF16EqualsString(str.Value, p.renamer.NameForSymbol(id.Ref)) {
								if property.DefaultValue != nil {
									p.printSpace()
									p.print("=")
									p.printSpace()
									p.printExpr(*property.DefaultValue, js_ast.LComma, 0)
								}
								continue
							}
						} else {
							p.printExpr(property.Key, js_ast.LLowest, 0)
						}
					} else {
						p.printExpr(property.Key, js_ast.LLowest, 0)
					}

					p.print(":")
					p.printSpace()
				}
				p.printBinding(property.Value)

				if property.DefaultValue != nil {
					p.printSpace()
					p.print("=")
					p.printSpace()
					p.printExpr(*property.DefaultValue, js_ast.LComma, 0)
				}
			}

			if !b.IsSingleLine {
				p.options.Indent--
				p.printNewline()
				p.printIndent()
			}
		}
		p.print("}")

	default:
		panic(fmt.Sprintf("Unexpected binding of type %T", binding.Data))
	}
}

func (p *printer) printSpace() {
	if !p.options.RemoveWhitespace {
		p.print(" ")
	}
}

func (p *printer) printNewline() {
	if !p.options.RemoveWhitespace {
		p.print("\n")
	}
}

func (p *printer) printSpaceBeforeOperator(next js_ast.OpCode) {
	if p.prevOpEnd == len(p.js) {
		prev := p.prevOp

"+ + y" => "+ +y" "+ ++ y" => "+ ++y" "x + + y" => "x+ +y" "x ++ + y" => "x+++y" "x + ++ y" => "x+ ++y" "-- >" => "-- >" "< ! --" => "<! --"

		if ((prev == js_ast.BinOpAdd || prev == js_ast.UnOpPos) && (next == js_ast.BinOpAdd || next == js_ast.UnOpPos || next == js_ast.UnOpPreInc)) ||
			((prev == js_ast.BinOpSub || prev == js_ast.UnOpNeg) && (next == js_ast.BinOpSub || next == js_ast.UnOpNeg || next == js_ast.UnOpPreDec)) ||
			(prev == js_ast.UnOpPostDec && next == js_ast.BinOpGt) ||
			(prev == js_ast.UnOpNot && next == js_ast.UnOpPreDec && len(p.js) > 1 && p.js[len(p.js)-2] == '<') {
			p.print(" ")
		}
	}
}

func (p *printer) printSemicolonAfterStatement() {
	if !p.options.RemoveWhitespace {
		p.print(";\n")
	} else {
		p.needsSemicolon = true
	}
}

func (p *printer) printSemicolonIfNeeded() {
	if p.needsSemicolon {
		p.print(";")
		p.needsSemicolon = false
	}
}

func (p *printer) printSpaceBeforeIdentifier() {
	buffer := p.js
	n := len(buffer)
	if n > 0 && (js_lexer.IsIdentifierContinue(rune(buffer[n-1])) || n == p.prevRegExpEnd) {
		p.print(" ")
	}
}

func (p *printer) printFnArgs(args []js_ast.Arg, hasRestArg bool, isArrow bool) {
	wrap := true

Minify "(a) => {}" as "a=>{}"

	if p.options.RemoveWhitespace && !hasRestArg && isArrow && len(args) == 1 {
		if _, ok := args[0].Binding.Data.(*js_ast.BIdentifier); ok && args[0].Default == nil {
			wrap = false
		}
	}

	if wrap {
		p.print("(")
	}

	for i, arg := range args {
		if i != 0 {
			p.print(",")
			p.printSpace()
		}
		if hasRestArg && i+1 == len(args) {
			p.print("...")
		}
		p.printBinding(arg.Binding)

		if arg.Default != nil {
			p.printSpace()
			p.print("=")
			p.printSpace()
			p.printExpr(*arg.Default, js_ast.LComma, 0)
		}
	}

	if wrap {
		p.print(")")
	}
}

func (p *printer) printFn(fn js_ast.Fn) {
	p.printFnArgs(fn.Args, fn.HasRestArg, false /* isArrow */)
	p.printSpace()
	p.printBlock(fn.Body.Loc, fn.Body.Stmts)
}

func (p *printer) printClass(class js_ast.Class) {
	if class.Extends != nil {
		p.print(" extends")
		p.printSpace()
		p.printExpr(*class.Extends, js_ast.LNew-1, 0)
	}
	p.printSpace()

	p.addSourceMapping(class.BodyLoc)
	p.print("{")
	p.printNewline()
	p.options.Indent++

	for _, item := range class.Properties {
		p.printSemicolonIfNeeded()
		p.printIndent()
		p.printProperty(item)

Need semicolons after class fields

		if item.Value == nil {
			p.printSemicolonAfterStatement()
		} else {
			p.printNewline()
		}
	}

	p.needsSemicolon = false
	p.options.Indent--
	p.printIndent()
	p.print("}")
}

func (p *printer) printProperty(item js_ast.Property) {
	if item.Kind == js_ast.PropertySpread {
		p.print("...")
		p.printExpr(*item.Value, js_ast.LComma, 0)
		return
	}

	if item.IsStatic {
		p.print("static")
		p.printSpace()
	}

	switch item.Kind {
	case js_ast.PropertyGet:
		p.printSpaceBeforeIdentifier()
		p.print("get")
		p.printSpace()

	case js_ast.PropertySet:
		p.printSpaceBeforeIdentifier()
		p.print("set")
		p.printSpace()
	}

	if item.Value != nil {
		if fn, ok := item.Value.Data.(*js_ast.EFunction); item.IsMethod && ok {
			if fn.Fn.IsAsync {
				p.printSpaceBeforeIdentifier()
				p.print("async")
				p.printSpace()
			}
			if fn.Fn.IsGenerator {
				p.print("*")
			}
		}
	}

	if item.IsComputed {
		p.print("[")
		p.printExpr(item.Key, js_ast.LComma, 0)
		p.print("]")

		if item.Value != nil {
			if fn, ok := item.Value.Data.(*js_ast.EFunction); item.IsMethod && ok {
				p.printFn(fn.Fn)
				return
			}

			p.print(":")
			p.printSpace()
			p.printExpr(*item.Value, js_ast.LComma, 0)
		}

		if item.Initializer != nil {
			p.printSpace()
			p.print("=")
			p.printSpace()
			p.printExpr(*item.Initializer, js_ast.LComma, 0)
		}
		return
	}

	switch key := item.Key.Data.(type) {
	case *js_ast.EPrivateIdentifier:
		p.printSymbol(key.Ref)

	case *js_ast.EString:
		p.addSourceMapping(item.Key.Loc)
		if p.canPrintIdentifierUTF16(key.Value) {
			p.printSpaceBeforeIdentifier()
			p.printIdentifierUTF16(key.Value)

Use a shorthand property if the names are the same

			if !p.options.UnsupportedFeatures.Has(compat.ObjectExtensions) && item.Value != nil {
				switch e := item.Value.Data.(type) {
				case *js_ast.EIdentifier:
					if js_lexer.UTF16EqualsString(key.Value, p.renamer.NameForSymbol(e.Ref)) {
						if item.Initializer != nil {
							p.printSpace()
							p.print("=")
							p.printSpace()
							p.printExpr(*item.Initializer, js_ast.LComma, 0)
						}
						return
					}

Make sure we're not using a property access instead of an identifier

					ref := js_ast.FollowSymbols(p.symbols, e.Ref)
					symbol := p.symbols.Get(ref)
					if symbol.NamespaceAlias == nil && js_lexer.UTF16EqualsString(key.Value, p.renamer.NameForSymbol(e.Ref)) {
						if item.Initializer != nil {
							p.printSpace()
							p.print("=")
							p.printSpace()
							p.printExpr(*item.Initializer, js_ast.LComma, 0)
						}
						return
					}
				}
			}
		} else {
			c := p.bestQuoteCharForString(key.Value, false /* allowBacktick */)
			p.print(c)
			p.printQuotedUTF16(key.Value, rune(c[0]))
			p.print(c)
		}

	default:
		p.printExpr(item.Key, js_ast.LLowest, 0)
	}

	if item.Kind != js_ast.PropertyNormal {
		f, ok := item.Value.Data.(*js_ast.EFunction)
		if ok {
			p.printFn(f.Fn)
			return
		}
	}

	if item.Value != nil {
		if fn, ok := item.Value.Data.(*js_ast.EFunction); item.IsMethod && ok {
			p.printFn(fn.Fn)
			return
		}

		p.print(":")
		p.printSpace()
		p.printExpr(*item.Value, js_ast.LComma, 0)
	}

	if item.Initializer != nil {
		p.printSpace()
		p.print("=")
		p.printSpace()
		p.printExpr(*item.Initializer, js_ast.LComma, 0)
	}
}

func (p *printer) bestQuoteCharForString(data []uint16, allowBacktick bool) string {
	if p.options.UnsupportedFeatures.Has(compat.TemplateLiteral) {
		allowBacktick = false
	}

	singleCost := 0
	doubleCost := 0
	backtickCost := 0

	for i, c := range data {
		switch c {
		case '\n':

The backslash for the newline costs an extra character for old-style string literals when compared to a template literal

				backtickCost--
			}
		case '\'':
			singleCost++
		case '"':
			doubleCost++
		case '`':
			backtickCost++

"${" sequences need to be escaped in template literals

			if i+1 < len(data) && data[i+1] == '{' {
				backtickCost++
			}
		}
	}

	c := "\""
	if doubleCost > singleCost {
		c = "'"
		if singleCost > backtickCost && allowBacktick {
			c = "`"
		}
	} else if doubleCost > backtickCost && allowBacktick {
		c = "`"
	}
	return c
}

type requireCallArgs struct {
	isES6Import       bool
	mustReturnPromise bool
}

func (p *printer) printRequireOrImportExpr(importRecordIndex uint32, leadingInteriorComments []js_ast.Comment) {
	record := &p.importRecords[importRecordIndex]
	p.printSpaceBeforeIdentifier()

Preserve "import()" expressions that don't point inside the bundle

	if !record.SourceIndex.IsValid() && record.Kind == ast.ImportDynamic && p.options.OutputFormat.KeepES6ImportExportSyntax() {
		p.print("import(")
		if len(leadingInteriorComments) > 0 {
			p.printNewline()
			p.options.Indent++
			for _, comment := range leadingInteriorComments {
				p.printIndentedComment(comment.Text)
			}
			p.printIndent()
		}
		p.printQuotedUTF8(record.Path.Text, true /* allowBacktick */)
		if len(leadingInteriorComments) > 0 {
			p.printNewline()
			p.options.Indent--
			p.printIndent()
		}
		p.print(")")
		return
	}

Make sure "import()" expressions return promises

	if record.Kind == ast.ImportDynamic {
		if p.options.UnsupportedFeatures.Has(compat.Arrow) {
			p.print("Promise.resolve().then(function()")
			p.printSpace()
			p.print("{")
			p.printNewline()
			p.options.Indent++
			p.printIndent()
			p.print("return ")
		} else {
			if p.options.RemoveWhitespace {
				p.print("Promise.resolve().then(()=>")
			} else {
				p.print("Promise.resolve().then(() => ")
			}
		}
	}

Make sure CommonJS imports are converted to ES6 if necessary

	if record.WrapWithToModule {
		p.printSymbol(p.options.ToModuleRef)
		p.print("(")
	}

If this import points inside the bundle, then call the "require()" function for that module directly. The linker must ensure that the module's require function exists by this point. Otherwise, fall back to a bare "require()" call. Then it's up to the user to provide it.

	if record.SourceIndex.IsValid() {
		p.printSymbol(p.options.WrapperRefForSource(record.SourceIndex.GetIndex()))
		p.print("()")
	} else {
		p.print("require(")
		p.printQuotedUTF8(record.Path.Text, true /* allowBacktick */)
		p.print(")")
	}

	if record.WrapWithToModule {
		p.print(")")
	}

	if record.Kind == ast.ImportDynamic {
		if p.options.UnsupportedFeatures.Has(compat.Arrow) {
			if !p.options.RemoveWhitespace {
				p.print(";")
			}
			p.printNewline()
			p.options.Indent--
			p.printIndent()
			p.print("})")
		} else {
			p.print(")")
		}
	}
}

const (
	forbidCall = 1 << iota
	forbidIn
	hasNonOptionalChainParent
)

func (p *printer) printUndefined(level js_ast.L) {
	if level >= js_ast.LPrefix {
		p.print("(void 0)")
	} else {
		p.printSpaceBeforeIdentifier()
		p.print("void 0")
		p.prevNumEnd = len(p.js)
	}
}

func (p *printer) printExpr(expr js_ast.Expr, level js_ast.L, flags int) {
	p.addSourceMapping(expr.Loc)

	switch e := expr.Data.(type) {
	case *js_ast.EMissing:

	case *js_ast.EUndefined:
		p.printUndefined(level)

	case *js_ast.ESuper:
		p.printSpaceBeforeIdentifier()
		p.print("super")

	case *js_ast.ENull:
		p.printSpaceBeforeIdentifier()
		p.print("null")

	case *js_ast.EThis:
		p.printSpaceBeforeIdentifier()
		p.print("this")

	case *js_ast.ESpread:
		p.print("...")
		p.printExpr(e.Value, js_ast.LComma, 0)

	case *js_ast.ENewTarget:
		p.printSpaceBeforeIdentifier()
		p.print("new.target")

	case *js_ast.EImportMeta:
		p.printSpaceBeforeIdentifier()
		p.print("import.meta")

	case *js_ast.ENew:
		wrap := level >= js_ast.LCall

		hasPureComment := !p.options.RemoveWhitespace && e.CanBeUnwrappedIfUnused
		if hasPureComment && level >= js_ast.LPostfix {
			wrap = true
		}

		if wrap {
			p.print("(")
		}

		if hasPureComment {
			p.print("/* @__PURE__ */ ")
		}

		p.printSpaceBeforeIdentifier()
		p.print("new")
		p.printSpace()
		p.printExpr(e.Target, js_ast.LNew, forbidCall)

Omit the "()" when minifying, but only when safe to do so

		if !p.options.RemoveWhitespace || len(e.Args) > 0 || level >= js_ast.LPostfix {
			p.print("(")
			for i, arg := range e.Args {
				if i != 0 {
					p.print(",")
					p.printSpace()
				}
				p.printExpr(arg, js_ast.LComma, 0)
			}
			p.print(")")
		}

		if wrap {
			p.print(")")
		}

	case *js_ast.ECall:
		wrap := level >= js_ast.LNew || (flags&forbidCall) != 0
		targetFlags := 0
		if e.OptionalChain == js_ast.OptionalChainNone {
			targetFlags = hasNonOptionalChainParent
		} else if (flags & hasNonOptionalChainParent) != 0 {
			wrap = true
		}

		hasPureComment := !p.options.RemoveWhitespace && e.CanBeUnwrappedIfUnused
		if hasPureComment && level >= js_ast.LPostfix {
			wrap = true
		}

		if wrap {
			p.print("(")
		}

		if hasPureComment {
			wasStmtStart := p.stmtStart == len(p.js)
			p.print("/* @__PURE__ */ ")
			if wasStmtStart {
				p.stmtStart = len(p.js)
			}
		}

We don't ever want to accidentally generate a direct eval expression here

		p.callTarget = e.Target.Data
		if !e.IsDirectEval && p.isUnboundEvalIdentifier(e.Target) {
			if p.options.RemoveWhitespace {
				p.print("(0,")
			} else {
				p.print("(0, ")
			}
			p.printExpr(e.Target, js_ast.LPostfix, 0)
			p.print(")")
		} else {
			p.printExpr(e.Target, js_ast.LPostfix, targetFlags)
		}

		if e.OptionalChain == js_ast.OptionalChainStart {
			p.print("?.")
		}
		p.print("(")
		for i, arg := range e.Args {
			if i != 0 {
				p.print(",")
				p.printSpace()
			}
			p.printExpr(arg, js_ast.LComma, 0)
		}
		p.print(")")
		if wrap {
			p.print(")")
		}

	case *js_ast.ERequire:
		wrap := level >= js_ast.LNew || (flags&forbidCall) != 0
		if wrap {
			p.print("(")
		}
		p.printRequireOrImportExpr(e.ImportRecordIndex, nil)
		if wrap {
			p.print(")")
		}

	case *js_ast.ERequireResolve:
		wrap := level >= js_ast.LNew || (flags&forbidCall) != 0
		if wrap {
			p.print("(")
		}
		p.printSpaceBeforeIdentifier()
		p.print("require.resolve(")
		p.printQuotedUTF8(p.importRecords[e.ImportRecordIndex].Path.Text, true /* allowBacktick */)
		p.print(")")
		if wrap {
			p.print(")")
		}

	case *js_ast.EImport:
		wrap := level >= js_ast.LNew || (flags&forbidCall) != 0
		if wrap {
			p.print("(")
		}

		var leadingInteriorComments []js_ast.Comment
		if !p.options.RemoveWhitespace {
			leadingInteriorComments = e.LeadingInteriorComments
		}

		if e.ImportRecordIndex.IsValid() {
			p.printRequireOrImportExpr(e.ImportRecordIndex.GetIndex(), leadingInteriorComments)

Handle non-string expressions

			p.printSpaceBeforeIdentifier()
			p.print("import(")
			if len(leadingInteriorComments) > 0 {
				p.printNewline()
				p.options.Indent++
				for _, comment := range e.LeadingInteriorComments {
					p.printIndentedComment(comment.Text)
				}
				p.printIndent()
			}
			p.printExpr(e.Expr, js_ast.LComma, 0)
			if len(leadingInteriorComments) > 0 {
				p.printNewline()
				p.options.Indent--
				p.printIndent()
			}
			p.print(")")
		}

		if wrap {
			p.print(")")
		}

	case *js_ast.EDot:
		wrap := false
		if e.OptionalChain == js_ast.OptionalChainNone {
			flags |= hasNonOptionalChainParent
		} else {
			if (flags & hasNonOptionalChainParent) != 0 {
				wrap = true
				p.print("(")
			}
			flags &= ^hasNonOptionalChainParent
		}
		p.printExpr(e.Target, js_ast.LPostfix, flags)
		if e.OptionalChain == js_ast.OptionalChainStart {
			p.print("?")
		}
		if p.canPrintIdentifier(e.Name) {

"1.toString" is a syntax error, so print "1 .toString" instead

				p.print(" ")
			}
			p.print(".")
			p.addSourceMapping(e.NameLoc)
			p.printIdentifier(e.Name)
		} else {
			p.print("[")
			p.addSourceMapping(e.NameLoc)
			p.printQuotedUTF8(e.Name, true /* allowBacktick */)
			p.print("]")
		}
		if wrap {
			p.print(")")
		}

	case *js_ast.EIndex:
		wrap := false
		if e.OptionalChain == js_ast.OptionalChainNone {
			flags |= hasNonOptionalChainParent
		} else {
			if (flags & hasNonOptionalChainParent) != 0 {
				wrap = true
				p.print("(")
			}
			flags &= ^hasNonOptionalChainParent
		}
		p.printExpr(e.Target, js_ast.LPostfix, flags)
		if e.OptionalChain == js_ast.OptionalChainStart {
			p.print("?.")
		}
		if private, ok := e.Index.Data.(*js_ast.EPrivateIdentifier); ok {
			if e.OptionalChain != js_ast.OptionalChainStart {
				p.print(".")
			}
			p.printSymbol(private.Ref)
		} else {
			p.print("[")
			p.printExpr(e.Index, js_ast.LLowest, 0)
			p.print("]")
		}
		if wrap {
			p.print(")")
		}

	case *js_ast.EIf:
		wrap := level >= js_ast.LConditional
		if wrap {
			p.print("(")
			flags &= ^forbidIn
		}
		p.printExpr(e.Test, js_ast.LConditional, flags&forbidIn)
		p.printSpace()
		p.print("?")
		p.printSpace()
		p.printExpr(e.Yes, js_ast.LYield, 0)
		p.printSpace()
		p.print(":")
		p.printSpace()
		p.printExpr(e.No, js_ast.LYield, flags&forbidIn)
		if wrap {
			p.print(")")
		}

	case *js_ast.EArrow:
		wrap := level >= js_ast.LAssign

		if wrap {
			p.print("(")
		}
		if e.IsAsync {
			p.printSpaceBeforeIdentifier()
			p.print("async")
			p.printSpace()
		}

		p.printFnArgs(e.Args, e.HasRestArg, true /* isArrow */)
		p.printSpace()
		p.print("=>")
		p.printSpace()

		wasPrinted := false
		if len(e.Body.Stmts) == 1 && e.PreferExpr {
			if s, ok := e.Body.Stmts[0].Data.(*js_ast.SReturn); ok && s.Value != nil {
				p.arrowExprStart = len(p.js)
				p.printExpr(*s.Value, js_ast.LComma, 0)
				wasPrinted = true
			}
		}
		if !wasPrinted {
			p.printBlock(e.Body.Loc, e.Body.Stmts)
		}
		if wrap {
			p.print(")")
		}

	case *js_ast.EFunction:
		n := len(p.js)
		wrap := p.stmtStart == n || p.exportDefaultStart == n
		if wrap {
			p.print("(")
		}
		p.printSpaceBeforeIdentifier()
		if e.Fn.IsAsync {
			p.print("async ")
		}
		p.print("function")
		if e.Fn.IsGenerator {
			p.print("*")
			p.printSpace()
		}
		if e.Fn.Name != nil {
			p.printSymbol(e.Fn.Name.Ref)
		}
		p.printFn(e.Fn)
		if wrap {
			p.print(")")
		}

	case *js_ast.EClass:
		n := len(p.js)
		wrap := p.stmtStart == n || p.exportDefaultStart == n
		if wrap {
			p.print("(")
		}
		p.printSpaceBeforeIdentifier()
		p.print("class")
		if e.Class.Name != nil {
			p.printSymbol(e.Class.Name.Ref)
		}
		p.printClass(e.Class)
		if wrap {
			p.print(")")
		}

	case *js_ast.EArray:
		p.print("[")
		if len(e.Items) > 0 {
			if !e.IsSingleLine {
				p.options.Indent++
			}

			for i, item := range e.Items {
				if i != 0 {
					p.print(",")
					if e.IsSingleLine {
						p.printSpace()
					}
				}
				if !e.IsSingleLine {
					p.printNewline()
					p.printIndent()
				}
				p.printExpr(item, js_ast.LComma, 0)

Make sure there's a comma after trailing missing items

				_, ok := item.Data.(*js_ast.EMissing)
				if ok && i == len(e.Items)-1 {
					p.print(",")
				}
			}

			if !e.IsSingleLine {
				p.options.Indent--
				p.printNewline()
				p.printIndent()
			}
		}
		p.print("]")

	case *js_ast.EObject:
		n := len(p.js)
		wrap := p.stmtStart == n || p.arrowExprStart == n
		if wrap {
			p.print("(")
		}
		p.print("{")
		if len(e.Properties) != 0 {
			if !e.IsSingleLine {
				p.options.Indent++
			}

			for i, item := range e.Properties {
				if i != 0 {
					p.print(",")
					if e.IsSingleLine {
						p.printSpace()
					}
				}
				if !e.IsSingleLine {
					p.printNewline()
					p.printIndent()
				}
				p.printProperty(item)
			}

			if !e.IsSingleLine {
				p.options.Indent--
				p.printNewline()
				p.printIndent()
			}
		}
		p.print("}")
		if wrap {
			p.print(")")
		}

	case *js_ast.EBoolean:
		if p.options.MangleSyntax {
			if level >= js_ast.LPrefix {
				if e.Value {
					p.print("(!0)")
				} else {
					p.print("(!1)")
				}
			} else {
				if e.Value {
					p.print("!0")
				} else {
					p.print("!1")
				}
			}
		} else {
			p.printSpaceBeforeIdentifier()
			if e.Value {
				p.print("true")
			} else {
				p.print("false")
			}
		}

If this was originally a template literal, print it as one as long as we're not minifying

		if e.PreferTemplate && !p.options.MangleSyntax && !p.options.UnsupportedFeatures.Has(compat.TemplateLiteral) {
			p.print("`")
			p.printQuotedUTF16(e.Value, '`')
			p.print("`")
			return
		}

		c := p.bestQuoteCharForString(e.Value, true /* allowBacktick */)
		p.print(c)
		p.printQuotedUTF16(e.Value, rune(c[0]))
		p.print(c)

Convert no-substitution template literals into strings if it's smaller

		if p.options.MangleSyntax && e.Tag == nil && len(e.Parts) == 0 {
			c := p.bestQuoteCharForString(e.Head, true /* allowBacktick */)
			p.print(c)
			p.printQuotedUTF16(e.Head, rune(c[0]))
			p.print(c)
			return
		}

Optional chains are forbidden in template tags

			if js_ast.IsOptionalChain(*e.Tag) {
				p.print("(")
				p.printExpr(*e.Tag, js_ast.LLowest, 0)
				p.print(")")
			} else {
				p.printExpr(*e.Tag, js_ast.LPostfix, 0)
			}
		}
		p.print("`")
		if e.Tag != nil {
			p.print(e.HeadRaw)
		} else {
			p.printQuotedUTF16(e.Head, '`')
		}
		for _, part := range e.Parts {
			p.print("${")
			p.printExpr(part.Value, js_ast.LLowest, 0)
			p.print("}")
			if e.Tag != nil {
				p.print(part.TailRaw)
			} else {
				p.printQuotedUTF16(part.Tail, '`')
			}
		}
		p.print("`")

	case *js_ast.ERegExp:
		buffer := p.js
		n := len(buffer)

Avoid forming a single-line comment

		if n > 0 && buffer[n-1] == '/' {
			p.print(" ")
		}
		p.print(e.Value)

Need a space before the next identifier to avoid it turning into flags

		p.prevRegExpEnd = len(p.js)

	case *js_ast.EBigInt:
		p.printSpaceBeforeIdentifier()
		p.print(e.Value)
		p.print("n")

	case *js_ast.ENumber:
		value := e.Value
		absValue := math.Abs(value)

		if value != value {
			p.printSpaceBeforeIdentifier()
			p.print("NaN")
		} else if value == positiveInfinity {
			p.printSpaceBeforeIdentifier()
			p.print("Infinity")
		} else if value == negativeInfinity {
			if level >= js_ast.LPrefix {
				p.print("(-Infinity)")
			} else {
				p.printSpaceBeforeOperator(js_ast.UnOpNeg)
				p.print("-Infinity")
			}
		} else {
			if !math.Signbit(value) {
				p.printSpaceBeforeIdentifier()
				p.printNonNegativeFloat(absValue)

Remember the end of the latest number

				p.prevNumEnd = len(p.js)

Expressions such as "(-1).toString" need to wrap negative numbers. Instead of testing for "value < 0" we test for "signbit(value)" and "!isNaN(value)" because we need this to be true for "-0" and "-0 < 0" is false.

				p.print("(-")
				p.printNonNegativeFloat(absValue)
				p.print(")")
			} else {
				p.printSpaceBeforeOperator(js_ast.UnOpNeg)
				p.print("-")
				p.printNonNegativeFloat(absValue)

Remember the end of the latest number

				p.prevNumEnd = len(p.js)
			}
		}

	case *js_ast.EIdentifier:
		name := p.renamer.NameForSymbol(e.Ref)
		wrap := len(p.js) == p.forOfInitStart && name == "let"

		if wrap {
			p.print("(")
		}

		p.printSpaceBeforeIdentifier()
		p.printIdentifier(name)

		if wrap {
			p.print(")")
		}

Potentially use a property access instead of an identifier

		ref := js_ast.FollowSymbols(p.symbols, e.Ref)
		symbol := p.symbols.Get(ref)

		if symbol.ImportItemStatus == js_ast.ImportItemMissing {
			p.printUndefined(level)
		} else if symbol.NamespaceAlias != nil {
			wrap := p.callTarget == e && e.WasOriginallyIdentifier
			if wrap {
				if p.options.RemoveWhitespace {
					p.print("(0,")
				} else {
					p.print("(0, ")
				}
			}
			p.printSymbol(symbol.NamespaceAlias.NamespaceRef)
			alias := symbol.NamespaceAlias.Alias
			if p.canPrintIdentifier(alias) {
				p.print(".")
				p.printIdentifier(alias)
			} else {
				p.print("[")
				p.printQuotedUTF8(alias, true /* allowBacktick */)
				p.print("]")
			}
			if wrap {
				p.print(")")
			}
		} else {
			p.printSymbol(e.Ref)
		}

	case *js_ast.EAwait:
		wrap := level >= js_ast.LPrefix

		if wrap {
			p.print("(")
		}

		p.printSpaceBeforeIdentifier()
		p.print("await")
		p.printSpace()
		p.printExpr(e.Value, js_ast.LPrefix-1, 0)

		if wrap {
			p.print(")")
		}

	case *js_ast.EYield:
		wrap := level >= js_ast.LAssign

		if wrap {
			p.print("(")
		}

		p.printSpaceBeforeIdentifier()
		p.print("yield")

		if e.Value != nil {
			if e.IsStar {
				p.print("*")
			}
			p.printSpace()
			p.printExpr(*e.Value, js_ast.LYield, 0)
		}

		if wrap {
			p.print(")")
		}

	case *js_ast.EUnary:
		entry := js_ast.OpTable[e.Op]
		wrap := level >= entry.Level

		if wrap {
			p.print("(")
		}

		if !e.Op.IsPrefix() {
			p.printExpr(e.Value, js_ast.LPostfix-1, 0)
		}

		if entry.IsKeyword {
			p.printSpaceBeforeIdentifier()
			p.print(entry.Text)
			p.printSpace()
		} else {
			p.printSpaceBeforeOperator(e.Op)
			p.print(entry.Text)
			p.prevOp = e.Op
			p.prevOpEnd = len(p.js)
		}

		if e.Op.IsPrefix() {
			p.printExpr(e.Value, js_ast.LPrefix-1, 0)
		}

		if wrap {
			p.print(")")
		}

	case *js_ast.EBinary:
		entry := js_ast.OpTable[e.Op]
		wrap := level >= entry.Level || (e.Op == js_ast.BinOpIn && (flags&forbidIn) != 0)

Destructuring assignments must be parenthesized

		if n := len(p.js); p.stmtStart == n || p.arrowExprStart == n {
			if _, ok := e.Left.Data.(*js_ast.EObject); ok {
				wrap = true
			}
		}

		if wrap {
			p.print("(")
			flags &= ^forbidIn
		}

		leftLevel := entry.Level - 1
		rightLevel := entry.Level - 1

		if e.Op.IsRightAssociative() {
			leftLevel = entry.Level
		}
		if e.Op.IsLeftAssociative() {
			rightLevel = entry.Level
		}

		switch e.Op {

"??" can't directly contain "||" or "&&" without being wrapped in parentheses

			if left, ok := e.Left.Data.(*js_ast.EBinary); ok && (left.Op == js_ast.BinOpLogicalOr || left.Op == js_ast.BinOpLogicalAnd) {
				leftLevel = js_ast.LPrefix
			}
			if right, ok := e.Right.Data.(*js_ast.EBinary); ok && (right.Op == js_ast.BinOpLogicalOr || right.Op == js_ast.BinOpLogicalAnd) {
				rightLevel = js_ast.LPrefix
			}

"**" can't contain certain unary expressions

			if left, ok := e.Left.Data.(*js_ast.EUnary); ok && left.Op.UnaryAssignTarget() == js_ast.AssignTargetNone {
				leftLevel = js_ast.LCall
			} else if _, ok := e.Left.Data.(*js_ast.EAwait); ok {
				leftLevel = js_ast.LCall

Undefined is printed as "void 0"

				leftLevel = js_ast.LCall

Negative numbers are printed using a unary operator

				leftLevel = js_ast.LCall

When minifying, booleans are printed as "!0 and "!1"

				if _, ok := e.Left.Data.(*js_ast.EBoolean); ok {
					leftLevel = js_ast.LCall
				}
			}
		}

		p.printExpr(e.Left, leftLevel, flags&forbidIn)

		if e.Op != js_ast.BinOpComma {
			p.printSpace()
		}

		if entry.IsKeyword {
			p.printSpaceBeforeIdentifier()
			p.print(entry.Text)
		} else {
			p.printSpaceBeforeOperator(e.Op)
			p.print(entry.Text)
			p.prevOp = e.Op
			p.prevOpEnd = len(p.js)
		}

		p.printSpace()

		p.printExpr(e.Right, rightLevel, flags&forbidIn)

		if wrap {
			p.print(")")
		}

	default:
		panic(fmt.Sprintf("Unexpected expression of type %T", expr.Data))
	}
}

func (p *printer) isUnboundEvalIdentifier(value js_ast.Expr) bool {

Using the original name here is ok since unbound symbols are not renamed

		symbol := p.symbols.Get(js_ast.FollowSymbols(p.symbols, id.Ref))
		return symbol.Kind == js_ast.SymbolUnbound && symbol.OriginalName == "eval"
	}
	return false
}

Convert an integer to a byte slice without any allocations

func (p *printer) smallIntToBytes(n int) []byte {
	wasNegative := n < 0

This assumes that -math.MinInt isn't a problem. This is fine because these integers are floating-point exponents which never go up that high.

		n = -n
	}

	bytes := p.intToBytesBuffer[:]
	start := len(bytes)

Write out the number from the end to the front

	for {
		start--
		bytes[start] = '0' + byte(n%10)
		n /= 10
		if n == 0 {
			break
		}
	}

Stick a negative sign on the front if needed

	if wasNegative {
		start--
		bytes[start] = '-'
	}

	return bytes[start:]
}

func parseSmallInt(bytes []byte) int {
	wasNegative := bytes[0] == '-'
	if wasNegative {
		bytes = bytes[1:]
	}

Parse the integer without any error checking. This doesn't need to handle integer overflow because these integers are floating-point exponents which never go up that high.

	n := 0
	for _, c := range bytes {
		n = n*10 + int(c-'0')
	}

	if wasNegative {
		return -n
	}
	return n
}

We can avoid the slow call to strconv.FormatFloat() for integers less than 1000 because we know that exponential notation will always be longer than the integer representation. This is not the case for 1000 which is "1e3".

	if absValue < 1000 {
		if asInt := int64(absValue); absValue == float64(asInt) {
			p.printBytes(p.smallIntToBytes(int(asInt)))
			return
		}
	}

Format this number into a byte slice so we can mutate it in place without further reallocation

	result := []byte(strconv.FormatFloat(absValue, 'g', -1, 64))

Simplify the exponent "e+05" => "e5" "e-05" => "e-5"

	if e := bytes.LastIndexByte(result, 'e'); e != -1 {
		from := e + 1
		to := from

		switch result[from] {

Strip off the leading "+"

			from++

Skip past the leading "-"

			to++
			from++
		}

Strip off leading zeros

		for from < len(result) && result[from] == '0' {
			from++
		}

		result = append(result[:to], result[from:]...)
	}

	dot := bytes.IndexByte(result, '.')

Simplify numbers starting with "0."

		afterDot := 2

Strip off the leading zero when minifying "0.5" => ".5"

		if p.options.RemoveWhitespace {
			result = result[1:]
			afterDot--
		}

Try using an exponent "0.001" => "1e-3"

		if result[afterDot] == '0' {
			i := afterDot + 1
			for result[i] == '0' {
				i++
			}
			remaining := result[i:]
			exponent := p.smallIntToBytes(afterDot - i - len(remaining))

Only switch if it's actually shorter

			if len(result) > len(remaining)+1+len(exponent) {
				result = append(append(remaining, 'e'), exponent...)
			}
		}

Try to get rid of a "." and maybe also an "e"

		if e := bytes.LastIndexByte(result, 'e'); e != -1 {
			integer := result[:dot]
			fraction := result[dot+1 : e]
			exponent := parseSmallInt(result[e+1:]) - len(fraction)

Handle small exponents by appending zeros instead

"1.2e1" => "12" "1.2e2" => "120" "1.2e3" => "1200"

				if len(result) >= len(integer)+len(fraction)+exponent {
					result = append(integer, fraction...)
					for i := 0; i < exponent; i++ {
						result = append(result, '0')
					}
				}

"1.2e4" => "12e3"

				exponent := p.smallIntToBytes(exponent)
				if len(result) >= len(integer)+len(fraction)+1+len(exponent) {
					result = append(append(append(integer, fraction...), 'e'), exponent...)
				}
			}
		}

Simplify numbers ending with "0" by trying to use an exponent "1000" => "1e3"

		i := len(result) - 1
		for i > 0 && result[i-1] == '0' {
			i--
		}
		remaining := result[:i]
		exponent := p.smallIntToBytes(len(result) - i)

Only switch if it's actually shorter

		if len(result) > len(remaining)+1+len(exponent) {
			result = append(append(remaining, 'e'), exponent...)
		}
	}

	p.printBytes(result)
}

func (p *printer) printDeclStmt(isExport bool, keyword string, decls []js_ast.Decl) {
	p.printIndent()
	p.printSpaceBeforeIdentifier()
	if isExport {
		p.print("export ")
	}
	p.printDecls(keyword, decls, 0)
	p.printSemicolonAfterStatement()
}

func (p *printer) printForLoopInit(init js_ast.Stmt) {
	switch s := init.Data.(type) {
	case *js_ast.SExpr:
		p.printExpr(s.Value, js_ast.LLowest, forbidIn)
	case *js_ast.SLocal:
		switch s.Kind {
		case js_ast.LocalVar:
			p.printDecls("var", s.Decls, forbidIn)
		case js_ast.LocalLet:
			p.printDecls("let", s.Decls, forbidIn)
		case js_ast.LocalConst:
			p.printDecls("const", s.Decls, forbidIn)
		}
	default:
		panic("Internal error")
	}
}

func (p *printer) printDecls(keyword string, decls []js_ast.Decl, flags int) {
	p.print(keyword)
	p.printSpace()

	for i, decl := range decls {
		if i != 0 {
			p.print(",")
			p.printSpace()
		}
		p.printBinding(decl.Binding)

		if decl.Value != nil {
			p.printSpace()
			p.print("=")
			p.printSpace()
			p.printExpr(*decl.Value, js_ast.LComma, flags)
		}
	}
}

func (p *printer) printBody(body js_ast.Stmt) {
	if block, ok := body.Data.(*js_ast.SBlock); ok {
		p.printSpace()
		p.printBlock(body.Loc, block.Stmts)
		p.printNewline()
	} else {
		p.printNewline()
		p.options.Indent++
		p.printStmt(body)
		p.options.Indent--
	}
}

func (p *printer) printBlock(loc logger.Loc, stmts []js_ast.Stmt) {
	p.addSourceMapping(loc)
	p.print("{")
	p.printNewline()

	p.options.Indent++
	for _, stmt := range stmts {
		p.printSemicolonIfNeeded()
		p.printStmt(stmt)
	}
	p.options.Indent--
	p.needsSemicolon = false

	p.printIndent()
	p.print("}")
}

func wrapToAvoidAmbiguousElse(s js_ast.S) bool {
	for {
		switch current := s.(type) {
		case *js_ast.SIf:
			if current.No == nil {
				return true
			}
			s = current.No.Data

		case *js_ast.SFor:
			s = current.Body.Data

		case *js_ast.SForIn:
			s = current.Body.Data

		case *js_ast.SForOf:
			s = current.Body.Data

		case *js_ast.SWhile:
			s = current.Body.Data

		case *js_ast.SWith:
			s = current.Body.Data

		default:
			return false
		}
	}
}

func (p *printer) printIf(s *js_ast.SIf) {
	p.printSpaceBeforeIdentifier()
	p.print("if")
	p.printSpace()
	p.print("(")
	p.printExpr(s.Test, js_ast.LLowest, 0)
	p.print(")")

	if yes, ok := s.Yes.Data.(*js_ast.SBlock); ok {
		p.printSpace()
		p.printBlock(s.Yes.Loc, yes.Stmts)

		if s.No != nil {
			p.printSpace()
		} else {
			p.printNewline()
		}
	} else if wrapToAvoidAmbiguousElse(s.Yes.Data) {
		p.printSpace()
		p.print("{")
		p.printNewline()

		p.options.Indent++
		p.printStmt(s.Yes)
		p.options.Indent--
		p.needsSemicolon = false

		p.printIndent()
		p.print("}")

		if s.No != nil {
			p.printSpace()
		} else {
			p.printNewline()
		}
	} else {
		p.printNewline()
		p.options.Indent++
		p.printStmt(s.Yes)
		p.options.Indent--

		if s.No != nil {
			p.printIndent()
		}
	}

	if s.No != nil {
		p.printSemicolonIfNeeded()
		p.printSpaceBeforeIdentifier()
		p.print("else")

		if no, ok := s.No.Data.(*js_ast.SBlock); ok {
			p.printSpace()
			p.printBlock(s.No.Loc, no.Stmts)
			p.printNewline()
		} else if no, ok := s.No.Data.(*js_ast.SIf); ok {
			p.printIf(no)
		} else {
			p.printNewline()
			p.options.Indent++
			p.printStmt(*s.No)
			p.options.Indent--
		}
	}
}

func (p *printer) printIndentedComment(text string) {

Re-indent multi-line comments

		for {
			newline := strings.IndexByte(text, '\n')
			if newline == -1 {
				break
			}
			p.printIndent()
			p.print(text[:newline+1])
			text = text[newline+1:]
		}
		p.printIndent()
		p.print(text)
		p.printNewline()

Print a mandatory newline after single-line comments

		p.printIndent()
		p.print(text)
		p.print("\n")
	}
}

func (p *printer) printStmt(stmt js_ast.Stmt) {
	p.addSourceMapping(stmt.Loc)

	switch s := stmt.Data.(type) {
	case *js_ast.SComment:
		text := s.Text
		if p.options.ExtractComments {
			if p.extractedComments == nil {
				p.extractedComments = make(map[string]bool)
			}
			p.extractedComments[text] = true
			break
		}
		p.printIndentedComment(text)

	case *js_ast.SFunction:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		if s.IsExport {
			p.print("export ")
		}
		if s.Fn.IsAsync {
			p.print("async ")
		}
		p.print("function")
		if s.Fn.IsGenerator {
			p.print("*")
			p.printSpace()
		}
		p.printSymbol(s.Fn.Name.Ref)
		p.printFn(s.Fn)
		p.printNewline()

	case *js_ast.SClass:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		if s.IsExport {
			p.print("export ")
		}
		p.print("class")
		p.printSymbol(s.Class.Name.Ref)
		p.printClass(s.Class)
		p.printNewline()

	case *js_ast.SEmpty:
		p.printIndent()
		p.print(";")
		p.printNewline()

	case *js_ast.SExportDefault:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("export default")
		p.printSpace()

Functions and classes must be wrapped to avoid confusion with their statement forms

			p.exportDefaultStart = len(p.js)

			p.printExpr(*s.Value.Expr, js_ast.LComma, 0)
			p.printSemicolonAfterStatement()
			return
		}

		switch s2 := s.Value.Stmt.Data.(type) {
		case *js_ast.SFunction:
			p.printSpaceBeforeIdentifier()
			if s2.Fn.IsAsync {
				p.print("async ")
			}
			p.print("function")
			if s2.Fn.IsGenerator {
				p.print("*")
				p.printSpace()
			}
			if s2.Fn.Name != nil {
				p.printSymbol(s2.Fn.Name.Ref)
			}
			p.printFn(s2.Fn)
			p.printNewline()

		case *js_ast.SClass:
			p.printSpaceBeforeIdentifier()
			p.print("class")
			if s2.Class.Name != nil {
				p.printSymbol(s2.Class.Name.Ref)
			}
			p.printClass(s2.Class)
			p.printNewline()

		default:
			panic("Internal error")
		}

	case *js_ast.SExportStar:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("export")
		p.printSpace()
		p.print("*")
		p.printSpace()
		if s.Alias != nil {
			p.print("as")
			p.printSpace()
			p.printSpaceBeforeIdentifier()
			p.printIdentifier(s.Alias.OriginalName)
			p.printSpace()
			p.printSpaceBeforeIdentifier()
		}
		p.print("from")
		p.printSpace()
		p.printQuotedUTF8(p.importRecords[s.ImportRecordIndex].Path.Text, false /* allowBacktick */)
		p.printSemicolonAfterStatement()

	case *js_ast.SExportClause:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("export")
		p.printSpace()
		p.print("{")

		if !s.IsSingleLine {
			p.options.Indent++
		}

		for i, item := range s.Items {
			if i != 0 {
				p.print(",")
				if s.IsSingleLine {
					p.printSpace()
				}
			}

			if !s.IsSingleLine {
				p.printNewline()
				p.printIndent()
			}
			name := p.renamer.NameForSymbol(item.Name.Ref)
			p.printIdentifier(name)
			if name != item.Alias {
				p.print(" as ")
				p.printIdentifier(item.Alias)
			}
		}

		if !s.IsSingleLine {
			p.options.Indent--
			p.printNewline()
			p.printIndent()
		}

		p.print("}")
		p.printSemicolonAfterStatement()

	case *js_ast.SExportFrom:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("export")
		p.printSpace()
		p.print("{")

		if !s.IsSingleLine {
			p.options.Indent++
		}

		for i, item := range s.Items {
			if i != 0 {
				p.print(",")
				if s.IsSingleLine {
					p.printSpace()
				}
			}

			if !s.IsSingleLine {
				p.printNewline()
				p.printIndent()
			}
			p.printIdentifier(item.OriginalName)
			if item.OriginalName != item.Alias {
				p.print(" as ")
				p.printIdentifier(item.Alias)
			}
		}

		if !s.IsSingleLine {
			p.options.Indent--
			p.printNewline()
			p.printIndent()
		}

		p.print("}")
		p.printSpace()
		p.print("from")
		p.printSpace()
		p.printQuotedUTF8(p.importRecords[s.ImportRecordIndex].Path.Text, false /* allowBacktick */)
		p.printSemicolonAfterStatement()

	case *js_ast.SLocal:
		switch s.Kind {
		case js_ast.LocalConst:
			p.printDeclStmt(s.IsExport, "const", s.Decls)
		case js_ast.LocalLet:
			p.printDeclStmt(s.IsExport, "let", s.Decls)
		case js_ast.LocalVar:
			p.printDeclStmt(s.IsExport, "var", s.Decls)
		}

	case *js_ast.SIf:
		p.printIndent()
		p.printIf(s)

	case *js_ast.SDoWhile:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("do")
		if block, ok := s.Body.Data.(*js_ast.SBlock); ok {
			p.printSpace()
			p.printBlock(s.Body.Loc, block.Stmts)
			p.printSpace()
		} else {
			p.printNewline()
			p.options.Indent++
			p.printStmt(s.Body)
			p.printSemicolonIfNeeded()
			p.options.Indent--
			p.printIndent()
		}
		p.print("while")
		p.printSpace()
		p.print("(")
		p.printExpr(s.Test, js_ast.LLowest, 0)
		p.print(")")
		p.printSemicolonAfterStatement()

	case *js_ast.SForIn:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("for")
		p.printSpace()
		p.print("(")
		p.printForLoopInit(s.Init)
		p.printSpace()
		p.printSpaceBeforeIdentifier()
		p.print("in")
		p.printSpace()
		p.printExpr(s.Value, js_ast.LLowest, 0)
		p.print(")")
		p.printBody(s.Body)

	case *js_ast.SForOf:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("for")
		if s.IsAwait {
			p.print(" await")
		}
		p.printSpace()
		p.print("(")
		p.forOfInitStart = len(p.js)
		p.printForLoopInit(s.Init)
		p.printSpace()
		p.printSpaceBeforeIdentifier()
		p.print("of")
		p.printSpace()
		p.printExpr(s.Value, js_ast.LComma, 0)
		p.print(")")
		p.printBody(s.Body)

	case *js_ast.SWhile:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("while")
		p.printSpace()
		p.print("(")
		p.printExpr(s.Test, js_ast.LLowest, 0)
		p.print(")")
		p.printBody(s.Body)

	case *js_ast.SWith:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("with")
		p.printSpace()
		p.print("(")
		p.printExpr(s.Value, js_ast.LLowest, 0)
		p.print(")")
		p.printBody(s.Body)

	case *js_ast.SLabel:
		p.printIndent()
		p.printSymbol(s.Name.Ref)
		p.print(":")
		p.printBody(s.Stmt)

	case *js_ast.STry:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("try")
		p.printSpace()
		p.printBlock(s.BodyLoc, s.Body)

		if s.Catch != nil {
			p.printSpace()
			p.print("catch")
			if s.Catch.Binding != nil {
				p.printSpace()
				p.print("(")
				p.printBinding(*s.Catch.Binding)
				p.print(")")
			}
			p.printSpace()
			p.printBlock(s.Catch.Loc, s.Catch.Body)
		}

		if s.Finally != nil {
			p.printSpace()
			p.print("finally")
			p.printSpace()
			p.printBlock(s.Finally.Loc, s.Finally.Stmts)
		}

		p.printNewline()

	case *js_ast.SFor:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("for")
		p.printSpace()
		p.print("(")
		if s.Init != nil {
			p.printForLoopInit(*s.Init)
		}
		p.print(";")
		p.printSpace()
		if s.Test != nil {
			p.printExpr(*s.Test, js_ast.LLowest, 0)
		}
		p.print(";")
		p.printSpace()
		if s.Update != nil {
			p.printExpr(*s.Update, js_ast.LLowest, 0)
		}
		p.print(")")
		p.printBody(s.Body)

	case *js_ast.SSwitch:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("switch")
		p.printSpace()
		p.print("(")
		p.printExpr(s.Test, js_ast.LLowest, 0)
		p.print(")")
		p.printSpace()
		p.print("{")
		p.printNewline()
		p.options.Indent++

		for _, c := range s.Cases {
			p.printSemicolonIfNeeded()
			p.printIndent()

			if c.Value != nil {
				p.print("case")
				p.printSpace()
				p.printExpr(*c.Value, js_ast.LLogicalAnd, 0)
			} else {
				p.print("default")
			}
			p.print(":")

			if len(c.Body) == 1 {
				if block, ok := c.Body[0].Data.(*js_ast.SBlock); ok {
					p.printSpace()
					p.printBlock(c.Body[0].Loc, block.Stmts)
					p.printNewline()
					continue
				}
			}

			p.printNewline()
			p.options.Indent++
			for _, stmt := range c.Body {
				p.printSemicolonIfNeeded()
				p.printStmt(stmt)
			}
			p.options.Indent--
		}

		p.options.Indent--
		p.printIndent()
		p.print("}")
		p.printNewline()
		p.needsSemicolon = false

	case *js_ast.SImport:
		itemCount := 0

		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("import")
		p.printSpace()

		if s.DefaultName != nil {
			p.printSymbol(s.DefaultName.Ref)
			itemCount++
		}

		if s.Items != nil {
			if itemCount > 0 {
				p.print(",")
				p.printSpace()
			}

			p.print("{")
			if !s.IsSingleLine {
				p.options.Indent++
			}

			for i, item := range *s.Items {
				if i != 0 {
					p.print(",")
					if s.IsSingleLine {
						p.printSpace()
					}
				}

				if !s.IsSingleLine {
					p.printNewline()
					p.printIndent()
				}
				p.printIdentifier(item.Alias)
				name := p.renamer.NameForSymbol(item.Name.Ref)
				if name != item.Alias {
					p.print(" as ")
					p.printIdentifier(name)
				}
			}

			if !s.IsSingleLine {
				p.options.Indent--
				p.printNewline()
				p.printIndent()
			}
			p.print("}")
			itemCount++
		}

		if s.StarNameLoc != nil {
			if itemCount > 0 {
				p.print(",")
				p.printSpace()
			}

			p.print("*")
			p.printSpace()
			p.print("as ")
			p.printSymbol(s.NamespaceRef)
			itemCount++
		}

		if itemCount > 0 {
			p.printSpace()
			p.printSpaceBeforeIdentifier()
			p.print("from")
			p.printSpace()
		}

		p.printQuotedUTF8(p.importRecords[s.ImportRecordIndex].Path.Text, false /* allowBacktick */)
		p.printSemicolonAfterStatement()

	case *js_ast.SBlock:
		p.printIndent()
		p.printBlock(stmt.Loc, s.Stmts)
		p.printNewline()

	case *js_ast.SDebugger:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("debugger")
		p.printSemicolonAfterStatement()

	case *js_ast.SDirective:
		c := p.bestQuoteCharForString(s.Value, false /* allowBacktick */)
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print(c)
		p.printQuotedUTF16(s.Value, rune(c[0]))
		p.print(c)
		p.printSemicolonAfterStatement()

	case *js_ast.SBreak:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("break")
		if s.Label != nil {
			p.print(" ")
			p.printSymbol(s.Label.Ref)
		}
		p.printSemicolonAfterStatement()

	case *js_ast.SContinue:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("continue")
		if s.Label != nil {
			p.print(" ")
			p.printSymbol(s.Label.Ref)
		}
		p.printSemicolonAfterStatement()

	case *js_ast.SReturn:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("return")
		if s.Value != nil {
			p.printSpace()
			p.printExpr(*s.Value, js_ast.LLowest, 0)
		}
		p.printSemicolonAfterStatement()

	case *js_ast.SThrow:
		p.printIndent()
		p.printSpaceBeforeIdentifier()
		p.print("throw")
		p.printSpace()
		p.printExpr(s.Value, js_ast.LLowest, 0)
		p.printSemicolonAfterStatement()

	case *js_ast.SExpr:
		p.printIndent()
		p.stmtStart = len(p.js)
		p.printExpr(s.Value, js_ast.LLowest, 0)
		p.printSemicolonAfterStatement()

	default:
		panic(fmt.Sprintf("Unexpected statement of type %T", stmt.Data))
	}
}

func (p *printer) shouldIgnoreSourceMap() bool {
	for _, c := range p.sourceMap {
		if c != ';' {
			return false
		}
	}
	return true
}

type Options struct {
	OutputFormat        config.Format
	RemoveWhitespace    bool
	MangleSyntax        bool
	ASCIIOnly           bool
	ExtractComments     bool
	AddSourceMappings   bool
	Indent              int
	ToModuleRef         js_ast.Ref
	WrapperRefForSource func(uint32) js_ast.Ref
	UnsupportedFeatures compat.JSFeature

If we're writing out a source map, this table of line start indices lets us do binary search on to figure out what line a given AST node came from

	LineOffsetTables []LineOffsetTable

This will be present if the input file had a source map. In that case we want to map all the way back to the original input file(s).

	InputSourceMap *sourcemap.SourceMap
}

type SourceMapChunk struct {
	Buffer []byte

This end state will be used to rewrite the start of the following source map chunk so that the delta-encoded VLQ numbers are preserved.

	EndState SourceMapState

There probably isn't a source mapping at the end of the file (nor should there be) but if we're appending another source map chunk after this one, we'll need to know how many characters were in the last line we generated.

	FinalGeneratedColumn int

	ShouldIgnore bool
}

type PrintResult struct {
	JS []byte

This source map chunk just contains the VLQ-encoded offsets for the "JS" field above. It's not a full source map. The bundler will be joining many source map chunks together to form the final source map.

	SourceMapChunk SourceMapChunk

	ExtractedComments map[string]bool
}

func Print(tree js_ast.AST, symbols js_ast.SymbolMap, r renamer.Renamer, options Options) PrintResult {
	p := &printer{
		symbols:            symbols,
		renamer:            r,
		importRecords:      tree.ImportRecords,
		options:            options,
		stmtStart:          -1,
		exportDefaultStart: -1,
		arrowExprStart:     -1,
		forOfInitStart:     -1,
		prevOpEnd:          -1,
		prevNumEnd:         -1,
		prevRegExpEnd:      -1,
		prevLoc:            logger.Loc{Start: -1},
		lineOffsetTables:   options.LineOffsetTables,

We automatically repeat the previous source mapping if we ever generate a line that doesn't start with a mapping. This helps give files more complete mapping coverage without gaps. However, we probably shouldn't do this if the input file has a nested source map that we will be remapping through. We have no idea what state that source map is in and it could be pretty scrambled. I've seen cases where blindly repeating the last mapping for subsequent lines gives very strange and unhelpful results with source maps from other tools.

		coverLinesWithoutMappings: options.InputSourceMap == nil,
	}

Add the top-level directive if present

	if tree.Directive != "" {
		p.printQuotedUTF8(tree.Directive, options.ASCIIOnly)
		p.print(";")
		p.printNewline()
	}

	for _, part := range tree.Parts {
		for _, stmt := range part.Stmts {
			p.printStmt(stmt)
			p.printSemicolonIfNeeded()
		}
	}

	p.updateGeneratedLineAndColumn()

	return PrintResult{
		JS:                p.js,
		ExtractedComments: p.extractedComments,
		SourceMapChunk: SourceMapChunk{
			Buffer:               p.sourceMap,
			EndState:             p.prevState,
			FinalGeneratedColumn: p.generatedColumn,
			ShouldIgnore:         p.shouldIgnoreSourceMap(),
		},
	}