1282 lines
62 KiB
TypeScript
1282 lines
62 KiB
TypeScript
import type { Access, AttributeNames, BaseFacts, BlockKind, CallArgument, CallFacts, ClassFacts, CommandContext, CommandFacts, CommandForm, CommandReading, CommandRule, CommandWord, CommonAttribute, DeclaratorFacts, Defines, DirectiveFacts, EmbeddedReading, EntryFacts, EnumFacts, ExceptionFacts, FactBlock, FunctionFacts, HealthFacts, ImportFacts, ImportItem, Inline, InstructionFlag, InterpreterFacts, IterationClause, IterationFacts, IterationResult, LanguageAdapter, Markdown, Node, OptionsFacts, ParameterFacts, PositionalParameter, PreviewTable, PropertyItem, PropertyRow, Redirect, Segment, SegmentGroup, SourceDocument, SourceRange, SourceSpan, StageFacts, StartFacts, TextBlock, TextTranslation, TokenRole, Translator, VariableBlock } from '../contract'
|
||
import { commonIndent, splitLines } from '../editText'
|
||
|
||
/** 按所在语法节点区分译法的关键字，例如 Python 的 else：if 后「否则」，for / while 后「未中断时」。 */
|
||
export interface ContextualWord {
|
||
/** 没有匹配的语境时用的译法。 */
|
||
zh: string
|
||
/**
|
||
* 语境 → 译法。语境是关键字节点往上的节点类型链，外层在前、用 ` > ` 连接：
|
||
* `raise_statement` 指父节点，`for_statement > else_clause` 指父节点 else_clause 且祖父节点 for_statement。
|
||
* 同时匹配多条时取最长的那条。
|
||
*/
|
||
byContext: Record<string, string>
|
||
}
|
||
|
||
/**
|
||
* 节点模板的一项。字符串是插入的译词，按关键字着色；另外两种是子节点槽位：
|
||
* `{ field }` 按 tree-sitter 字段名取子节点，`{ namedChild }` 按具名子节点序号（0 起，不计注释）取，给语法没定义字段名的节点用。
|
||
*/
|
||
export type TemplatePart = string | { field: string } | { namedChild: number }
|
||
|
||
/** 一种语言的中文直译表。引擎替换表里的关键字、内置类型、运算符，按节点模板重排少数语法结构，其余原样照抄。 */
|
||
/**
|
||
* 一种语言的中文词表：只有叫法和语法节点类型，语言知道的结构（词性、遍历）翻译时从 SourceDocument.language 取——
|
||
* 未翻译的词、关键字译文按它给的词性着色，它认出的遍历整个换成一个框，框里按读的顺序重排（见 iterationGroup）。
|
||
*/
|
||
export interface ZhTable {
|
||
id: string
|
||
label: string
|
||
languageIds: string[]
|
||
/** 关键字：按语法树里匿名节点的类型查（Python 的 `not in` 这类组合关键字本身也是一个匿名节点）；随所在语法节点换译法的写成 ContextualWord。 */
|
||
keywords: Record<string, string | ContextualWord>
|
||
/** 字面量关键字：具名的叶子节点（true / None …），按节点类型查。 */
|
||
literals: Record<string, string>
|
||
/** 内置类型：只在 isTypePosition 成立的叶子上按原文查。 */
|
||
builtinTypes: Record<string, string>
|
||
isTypePosition: (leaf: Node) => boolean
|
||
/** 翻成文字的运算符：只认父节点在 operatorParents 里的运算符节点（TS 的非空断言 `x!` 不算）。 */
|
||
operators: Record<string, string>
|
||
operatorParents: string[]
|
||
/** 空字面量（没有具名子节点）整个换成一个词：空数组 / 空列表 / 空字典 / 空对象。 */
|
||
emptyLiterals: Record<string, string>
|
||
/** 字典 / 对象字面量（语言插件给了 entries 的值）的叫法，按节点类型查：属性表里的值写成「字典 3 项」，各项展开成子行。 */
|
||
mapLiterals: Record<string, string>
|
||
/** 整段原样照抄、按字符串着色的节点。 */
|
||
stringNodes: string[]
|
||
/** 注释（或 docstring）节点去掉注释符号后的原文，可以多行；text 是节点的源码去掉代码阅读的记号（SourceDocument.markers）以后剩下的。 */
|
||
commentText: (node: Node, text: string) => string
|
||
/**
|
||
* 节点模板：节点类型 → 按中文语序重排的译文。命中的节点只输出模板里的译词和槽位，节点自己的关键字、标点不再输出；
|
||
* 槽位里的子节点照常递归翻译。模板引用的子节点缺了，或节点没整个落在要翻的这一段里，就不重排、照常翻。
|
||
*/
|
||
nodeTemplates: Record<string, TemplatePart[]>
|
||
/**
|
||
* 节点读法：节点类型 → 怎么读成一行译文，给模板排不了的写法（shell 的 `${x:-d}`、`[ -f x ]`、`case`）。
|
||
* 命中的节点只输出读法给的，不再照常翻它的子节点；读法返回 undefined 就照常翻。读法里要翻子节点用 ReaderContext.say。
|
||
*/
|
||
readers?: Record<string, NodeReader>
|
||
/** 并成一块的几条同种语句（TextBlock.group）怎么读：按第一条的节点类型查，读成一行字加一张表；没有的照常翻第一条。 */
|
||
groups?: Record<string, GroupReader>
|
||
/** 解释器开关（SwitchFacts.id）的叫法：errexit → 出错就停；没有的写源码里的写法。 */
|
||
switches?: Record<string, string>
|
||
/** 解析指令（ParserDirective.name）的叫法和补充说明：syntax → 语法（BuildKit 按这个版本读这份文件）；没有的写原名。 */
|
||
directives?: Record<string, { label: string, note?: string }>
|
||
}
|
||
|
||
/** 读一个节点时能用的（见 ZhTable.readers）。 */
|
||
export interface ReaderContext {
|
||
document: SourceDocument
|
||
/** 照常翻一个节点（照样认读法、命令），只翻落在要翻的这一段里的部分。 */
|
||
say: (node: Node) => Inline[]
|
||
/** node 整个落在要翻的这一段里：复合语句的块只翻头部，读法按它决定写不写主体里的东西。 */
|
||
whole: (node: Node) => boolean
|
||
/** 源码里一个名字照抄、按 role 着色（不给 role 按语言插件定的词性），名字带源码下标。 */
|
||
name: (node: Node, role?: TokenRole) => Segment
|
||
/** 读一条命令（命令规则读的，没有规则时命令名照抄、各个词照常翻），连重定向；不是命令时没有。 */
|
||
command: (node: Node) => CommandReading | undefined
|
||
}
|
||
|
||
/** 一种节点的读法：一行译文，可以带一张排在这一段下面的表（分情况的「情况 / 做什么」）；返回 undefined 就照常翻。 */
|
||
export type NodeReader = (node: Node, context: ReaderContext) => Inline[] | { text: Inline[], table?: PreviewTable } | undefined
|
||
|
||
/** 并成一块的几条同种语句的读法（见 ZhTable.groups）：照源码顺序的各条。 */
|
||
export type GroupReader = (nodes: Node[], context: ReaderContext) => { text: Inline[], table?: PreviewTable }
|
||
|
||
interface Token {
|
||
/** 注释是空文字：不进译文，只在前后留一个空格。 */
|
||
text: string
|
||
role: TokenRole
|
||
translated: boolean
|
||
/** 源码里的原文，用来认出逗号、括号；模板插入的译词、框没有原文。 */
|
||
raw: string
|
||
/** 前面要空一格：源码里它前面有空白，或者模板在它前面插了别的项。 */
|
||
gap: boolean
|
||
/** 照抄的名字、字符串在源码里的起始下标，见 Segment.offset。 */
|
||
offset?: number
|
||
/** 整个换成一个框的节点（写在表达式里的遍历）：输出这个框，text 为空。 */
|
||
group?: SegmentGroup
|
||
}
|
||
|
||
/**
|
||
* 排在这一段译文下面的那张表：跨几行写的调用的实参表（见 emitCall），或读法给的表（分情况的「情况 / 做什么」）。
|
||
* 一段只排一张，排了的记在 table 里。
|
||
*/
|
||
interface TableSlot {
|
||
table?: PreviewTable
|
||
}
|
||
|
||
/**
|
||
* 按译文顺序收集词：cursor 是上一个词在源码里的结尾，gap 是下一个词前面要不要空一格。
|
||
* slot 是这一段下面能不能排一张表（跨几行写的调用、读法给的表）；没有就照常翻。
|
||
*/
|
||
interface Collector {
|
||
document: SourceDocument
|
||
/** 要翻的那个节点 */
|
||
root: Node
|
||
tokens: Token[]
|
||
cursor: number
|
||
gap: boolean
|
||
slot?: TableSlot
|
||
}
|
||
|
||
const CONTEXT_SEPARATOR = ' > '
|
||
const CLOSING = new Set([')', ']', '}'])
|
||
/** 跟在一个式子后面、不属于这个式子的标点：右括号、逗号、分号、冒号。 */
|
||
const SEPARATORS = new Set([...CLOSING, ',', ';', ':'])
|
||
const HEADER_OPENERS = new Set(['{', ':'])
|
||
/** 源码着色里代表名字的词性：着这几种色的具名叶子是标识符、属性名、类型名，能查定义。 */
|
||
const NAME_ROLES = new Set<TokenRole>(['variable', 'constant', 'function', 'type'])
|
||
|
||
function firstLeaf(node: Node): Node {
|
||
const first = node.child(0)
|
||
return first ? firstLeaf(first) : node
|
||
}
|
||
|
||
function commentsWithin(node: Node, span: SourceSpan): Node[] {
|
||
if (node.endIndex <= span.startIndex || node.startIndex >= span.endIndex)
|
||
return []
|
||
if (node.type === 'comment')
|
||
return [node]
|
||
return node.children.flatMap(child => child ? commentsWithin(child, span) : [])
|
||
}
|
||
|
||
/** 节点的源码去掉整个落在它里面的记号（中文名记号 `@Comment 当前用户` 不是说明文字）。 */
|
||
function withoutMarkers(node: Node, markers: readonly SourceRange[]): string {
|
||
let text = ''
|
||
let cursor = node.startIndex
|
||
for (const marker of markers.filter(m => m.startIndex >= node.startIndex && m.endIndex <= node.endIndex).sort((a, b) => a.startIndex - b.startIndex)) {
|
||
text += node.text.slice(cursor - node.startIndex, marker.startIndex - node.startIndex)
|
||
cursor = Math.max(cursor, marker.endIndex)
|
||
}
|
||
return text + node.text.slice(cursor - node.startIndex)
|
||
}
|
||
|
||
function sameSpan(a: Node, b: Node): boolean {
|
||
return a.startIndex === b.startIndex && a.endIndex === b.endIndex && a.type === b.type
|
||
}
|
||
|
||
/**
|
||
* 在要翻的 root 里，node 后面还紧跟着包着它的式子的别的部分（`.model_dump()`、`[0]`、`or x`）：从 node 往上找第一个后面有东西的节点，
|
||
* 那个东西不是右括号、逗号、分号、冒号（语法之外的注释不算）。
|
||
*/
|
||
function continued(node: Node, root: Node): boolean {
|
||
for (let current: Node | null = node; current && !sameSpan(current, root); current = current.parent) {
|
||
let next = current.nextSibling
|
||
while (next?.isExtra)
|
||
next = next.nextSibling
|
||
if (next)
|
||
return !SEPARATORS.has(next.type)
|
||
}
|
||
return false
|
||
}
|
||
|
||
function isComment(token: Token): boolean {
|
||
return token.text === '' && !token.group
|
||
}
|
||
|
||
/** 下一个要输出的词的原文（跳过注释）。 */
|
||
function nextRaw(tokens: Token[], i: number): string {
|
||
for (let j = i + 1; j < tokens.length; j++) {
|
||
if (!isComment(tokens[j]))
|
||
return tokens[j].raw
|
||
}
|
||
return ''
|
||
}
|
||
|
||
function oneLine(text: string): string {
|
||
return text.replace(/\s*\n\s*/g, ' ')
|
||
}
|
||
|
||
/**
|
||
* 注释的 Markdown 原文：各个注释节点去掉注释符号后（texts）拼起来，去掉行尾空白、首尾空行和共同缩进，保留换行（段落之间的空行也留着）。
|
||
* 只有一个节点（块注释、docstring）时第一行紧跟在引号 / 注释符号后面，本来就没有缩进，不参与算共同缩进（同 PEP 257 的做法）；
|
||
* 几条行注释连成的一块每行地位相同，第一行也算，缩进四格的代码块才留得住。
|
||
*/
|
||
function docText(texts: string[]): string {
|
||
const lines = splitLines(texts.join('\n')).map(line => line.trimEnd())
|
||
const quoted = texts.length === 1
|
||
const indent = commonIndent(quoted ? lines.slice(1) : lines).length
|
||
return lines.map((line, i) => quoted && i === 0 ? line.trim() : line.slice(indent)).join('\n').replace(/^\n+|\n+$/g, '')
|
||
}
|
||
|
||
/** 模板槽位引用的子节点；注释是语法之外的节点，不占具名子节点序号。 */
|
||
function slotChild(node: Node, slot: Exclude<TemplatePart, string>): Node | null | undefined {
|
||
if ('field' in slot)
|
||
return node.childForFieldName(slot.field)
|
||
return node.namedChildren.filter(child => child && !child.isExtra)[slot.namedChild]
|
||
}
|
||
|
||
const ATTRIBUTE_ZH: Record<CommonAttribute, string> = { kind: '种类', access: '访问级别', async: '异步', abstract: '抽象', typeOnly: '仅类型' }
|
||
const KIND_ZH: Record<BlockKind, string> = {
|
||
import: '导入',
|
||
function: '函数',
|
||
class: '类',
|
||
enum: '枚举',
|
||
exception: '异常',
|
||
type: '类型',
|
||
field: '字段',
|
||
const: '常量',
|
||
variable: '变量',
|
||
if: '条件',
|
||
else: '否则',
|
||
loop: '循环',
|
||
try: '尝试',
|
||
catch: '捕获',
|
||
finally: '最终',
|
||
return: '返回',
|
||
statement: '语句',
|
||
comment: '注释',
|
||
interpreter: '解释器',
|
||
options: '选项',
|
||
case: '分情况',
|
||
command: '命令',
|
||
pipeline: '管道',
|
||
directive: '解析指令',
|
||
stage: '阶段',
|
||
run: '执行',
|
||
copy: '复制',
|
||
setting: '设置',
|
||
label: '标签',
|
||
health: '健康检查',
|
||
start: '启动',
|
||
}
|
||
const ACCESS_ZH: Record<Access, string> = { public: '公共', private: '私有', protected: '受保护' }
|
||
const ITERATION_ZH: Record<IterationResult, string> = { generator: '生成器', list: '列表', set: '集合', dict: '字典' }
|
||
/** 被调用的名字定义成这几种块时，调用就是新建对象。 */
|
||
const CONSTRUCTED = new Set<BlockKind | undefined>(['class', 'exception', 'enum'])
|
||
const ARGUMENT_COLUMNS = ['参数', '值']
|
||
const YES_NO: Record<'true' | 'false', string> = { true: '是', false: '否' }
|
||
|
||
/** 筛选属性和值的中文叫法；属性表里的访问级别、异步、抽象、仅类型这几行，属性名和值也用这一套。 */
|
||
const ZH_ATTRIBUTE_NAMES: AttributeNames = {
|
||
attributes: ATTRIBUTE_ZH,
|
||
values: { kind: KIND_ZH, access: ACCESS_ZH, async: YES_NO, abstract: YES_NO, typeOnly: YES_NO },
|
||
}
|
||
|
||
export function createZhTemplateTranslator(table: ZhTable): Translator {
|
||
const operatorParents = new Set(table.operatorParents)
|
||
const stringNodes = new Set(table.stringNodes)
|
||
const contextDepth = Math.max(0, ...Object.values(table.keywords).flatMap(word =>
|
||
typeof word === 'string' ? [] : Object.keys(word.byContext).map(context => context.split(CONTEXT_SEPARATOR).length)))
|
||
|
||
/** 从父节点往上逐层拼出语境去查，更长（更具体）的匹配覆盖更短的。 */
|
||
function keywordText(word: string | ContextualWord, node: Node): string {
|
||
if (typeof word === 'string')
|
||
return word
|
||
let text = word.zh
|
||
let context = ''
|
||
let ancestor = node.parent
|
||
for (let level = 0; ancestor && level < contextDepth; level++, ancestor = ancestor.parent) {
|
||
context = context ? `${ancestor.type}${CONTEXT_SEPARATOR}${context}` : ancestor.type
|
||
text = word.byContext[context] ?? text
|
||
}
|
||
return text
|
||
}
|
||
|
||
/** 照抄这个节点时带不带源码下标：名字（按名字着色的具名叶子）和整段字符串带，标点、数字、关键字不带。 */
|
||
function offsetOf(node: Node, language: LanguageAdapter): number | undefined {
|
||
const linkable = stringNodes.has(node.type) || (node.childCount === 0 && node.isNamed && NAME_ROLES.has(language.tokenRole(node)))
|
||
return linkable ? node.startIndex : undefined
|
||
}
|
||
|
||
/** 能整个换成中文的节点 → 译文和词性。 */
|
||
function translateNode(node: Node, language: LanguageAdapter): { text: string, role: TokenRole } | undefined {
|
||
if (node.childCount === 0 && table.isTypePosition(node)) {
|
||
const type = table.builtinTypes[node.text]
|
||
if (type)
|
||
return { text: type, role: 'type' }
|
||
}
|
||
const parent = node.parent
|
||
if (!node.isNamed) {
|
||
if (parent && operatorParents.has(parent.type) && table.operators[node.type])
|
||
return { text: table.operators[node.type], role: 'operator' }
|
||
const word = table.keywords[node.type]
|
||
if (word !== undefined)
|
||
return { text: keywordText(word, node), role: language.tokenRole(firstLeaf(node)) }
|
||
return undefined
|
||
}
|
||
if (node.childCount === 0 && table.literals[node.type])
|
||
return { text: table.literals[node.type], role: language.tokenRole(node) }
|
||
if (table.emptyLiterals[node.type] && node.namedChildCount === 0)
|
||
return { text: table.emptyLiterals[node.type], role: 'literal' }
|
||
return undefined
|
||
}
|
||
|
||
/** 节点命中模板时，模板各项换成译词或子节点；不该重排时返回 undefined。 */
|
||
function templateParts(node: Node, span: SourceSpan): (string | Node)[] | undefined {
|
||
const template = table.nodeTemplates[node.type]
|
||
if (!template || node.startIndex < span.startIndex || node.endIndex > span.endIndex)
|
||
return undefined
|
||
const parts = template.map(part => typeof part === 'string' ? part : slotChild(node, part))
|
||
return parts.every((part): part is string | Node => part != null) ? parts : undefined
|
||
}
|
||
|
||
/**
|
||
* 源码里上一个词到 index 之间有空白，下一个词前面就空一格；跨行写的一对圆括号挤成一行时贴着里面的内容
|
||
* （`f(\n x\n)` 写成 `f(x)`），方括号、花括号照旧空一格。
|
||
*/
|
||
function advance(out: Collector, index: number): void {
|
||
const between = out.document.source.slice(out.cursor, index)
|
||
const hugging = between.includes('\n') && (out.tokens.at(-1)?.raw === '(' || out.document.source[index] === ')')
|
||
out.gap ||= !hugging && /\s/.test(between)
|
||
out.cursor = index
|
||
}
|
||
|
||
function emit(out: Collector, token: Omit<Token, 'gap'>): void {
|
||
out.tokens.push({ ...token, gap: out.gap })
|
||
out.gap = false
|
||
}
|
||
|
||
function emitSource(out: Collector, node: Node, token: Omit<Token, 'gap' | 'raw'>): void {
|
||
advance(out, node.startIndex)
|
||
emit(out, { ...token, raw: node.text })
|
||
out.cursor = node.endIndex
|
||
}
|
||
|
||
/** 语言插件认出的写在表达式里的遍历；没整个落在要翻的这一段里就不重排、照常翻。 */
|
||
function iterationIn(node: Node, span: SourceSpan, language: LanguageAdapter): IterationFacts | undefined {
|
||
if (!language.iteration || node.startIndex < span.startIndex || node.endIndex > span.endIndex)
|
||
return undefined
|
||
return language.iteration(node)
|
||
}
|
||
|
||
/**
|
||
* 遍历按读的顺序重排进一个框：「遍历 <来源> 的每个 <变量>」，紧挨着的几节 for 用顿号接着写（异步的写「异步遍历」，
|
||
* 同步、异步换了就另起一句），条件在它所在的位置写「，仅当 <条件>」，最后「，取 <值>」，字典是「<键>: <值>」。
|
||
* 来源、变量、条件、值照常翻，里面的名字带源码下标，套着的遍历是框里的框。
|
||
*/
|
||
function iterationGroup(facts: IterationFacts, span: SourceSpan, document: SourceDocument): SegmentGroup {
|
||
const segments: Inline[] = []
|
||
const plain = (text: string): void => pushSegment(segments, text, 'plain')
|
||
const word = (text: string): void => pushSegment(segments, text, 'keyword')
|
||
const say = (node: Node): void => appendInline(segments, joinTokens(tokensIn(node, span, document)))
|
||
let previous: IterationClause | undefined
|
||
for (const clause of facts.clauses) {
|
||
if (clause.kind === 'if') {
|
||
plain('，')
|
||
word('仅当')
|
||
plain(' ')
|
||
say(clause.condition)
|
||
}
|
||
else {
|
||
if (previous?.kind === 'for' && previous.async === clause.async) {
|
||
plain('、')
|
||
}
|
||
else {
|
||
if (previous)
|
||
plain('，')
|
||
word(clause.async ? '异步遍历' : '遍历')
|
||
plain(' ')
|
||
}
|
||
say(clause.source)
|
||
plain(' ')
|
||
word('的每个')
|
||
plain(' ')
|
||
say(clause.variables)
|
||
}
|
||
previous = clause
|
||
}
|
||
plain('，')
|
||
word('取')
|
||
plain(' ')
|
||
if (facts.key) {
|
||
say(facts.key)
|
||
plain(': ')
|
||
}
|
||
say(facts.value)
|
||
return { style: 'iteration', label: ITERATION_ZH[facts.result], segments }
|
||
}
|
||
|
||
/**
|
||
* 遍历整个换成一个框；和外面的调用共用的那对括号照常写在框外，紧贴着框。
|
||
* 只翻这个遍历（它是调用的实参，排在实参表里）时外面的调用不在译文里，这对括号也不写。
|
||
*/
|
||
function emitIteration(out: Collector, node: Node, facts: IterationFacts, span: SourceSpan): void {
|
||
const alone = node.startIndex === out.root.startIndex && node.endIndex === out.root.endIndex
|
||
const shared = alone ? undefined : facts.sharedParentheses
|
||
const bracket = (paren: Node): Omit<Token, 'gap' | 'raw'> => ({ text: paren.text, role: out.document.language.tokenRole(paren), translated: false })
|
||
if (shared)
|
||
emitSource(out, shared.open, bracket(shared.open))
|
||
else
|
||
advance(out, node.startIndex)
|
||
emit(out, { text: '', role: 'plain', translated: true, raw: '', group: iterationGroup(facts, span, out.document) })
|
||
if (shared) {
|
||
out.cursor = shared.close.startIndex
|
||
emitSource(out, shared.close, bracket(shared.close))
|
||
}
|
||
out.cursor = node.endIndex
|
||
}
|
||
|
||
/** 要排成一行、实参摊成表的调用：实参列表跨了几行，整个落在要翻的这一段里，这一段还没排过别的调用。 */
|
||
function expandedCall(node: Node, span: SourceSpan, slot: TableSlot, { language: { call } }: SourceDocument): CallFacts | undefined {
|
||
if (slot.table || !call || node.startIndex < span.startIndex || node.endIndex > span.endIndex)
|
||
return undefined
|
||
const facts = call(node)
|
||
return facts && facts.argumentList.startPosition.row !== facts.argumentList.endPosition.row ? facts : undefined
|
||
}
|
||
|
||
/**
|
||
* node 落在 span 里的那一段翻成一行译文，里面跨几行写的第一个调用排成一行「新建 / 调用 …」，实参摊成表（见 emitCall）。
|
||
* 一行译文的块、常量 / 变量的值、实参表里每个值各翻各的，各自排自己里面的那个调用。
|
||
*/
|
||
function expandedText(node: Node, span: SourceSpan, document: SourceDocument): { text: Inline[], table?: PreviewTable } {
|
||
const slot: TableSlot = {}
|
||
const text = joinTokens(tokensIn(node, span, document, slot))
|
||
return { text, ...(slot.table && { table: slot.table }) }
|
||
}
|
||
|
||
/**
|
||
* 实参表：参数 | 值，每个实参一行，照源码顺序。按名字传的写名字，按位置传的写「第 N 个」（只数按位置传的），
|
||
* 展开的写「展开」（*args、...args）、「展开关键字」（**kwargs）。值照常翻；值里有跨几行写的调用时同样排成一行，
|
||
* 它的实参表排在值那一格里（PropertyItem.table），一层层往下套。
|
||
*/
|
||
function argumentTable(facts: CallFacts, span: SourceSpan, document: SourceDocument): PreviewTable {
|
||
let position = 0
|
||
const label = (arg: CallArgument): Inline[] => {
|
||
switch (arg.kind) {
|
||
case 'keyword':
|
||
return [segment(arg.name.text, 'variable', offsetOf(arg.name, document.language))]
|
||
case 'positional':
|
||
position++
|
||
return [hint(`第 ${position} 个`)]
|
||
case 'spread':
|
||
return [hint('展开')]
|
||
case 'keywordSpread':
|
||
return [hint('展开关键字')]
|
||
}
|
||
}
|
||
const row = (arg: CallArgument): PropertyItem => {
|
||
const { text, table } = expandedText(arg.value, span, document)
|
||
return { cells: [label(arg), text], ...(table && { table }) }
|
||
}
|
||
return { columns: ARGUMENT_COLUMNS, items: facts.arguments.map(row) }
|
||
}
|
||
|
||
/**
|
||
* 调用排成一行「新建 / 调用 被调用的名字 · N 个参数」，实参摊成表记在 slot 里。被调用的名字照常翻、带源码下标，
|
||
* 最后那个名字新建时按类型、调用时按函数着色，前面几节按变量着色。新建还是调用看写法（CallFacts.construct），
|
||
* 写法看不出来的问名字的定义：是类（含异常、枚举）就是新建，还没查到、查不到按调用算。
|
||
* 调用后面还紧跟着同一个式子的别的部分（见 continued）时整行括起来：「(新建 X · 3 个参数).model_dump()」。
|
||
*/
|
||
function emitCall(out: Collector, slot: TableSlot, node: Node, facts: CallFacts, span: SourceSpan): void {
|
||
slot.table = argumentTable(facts, span, out.document)
|
||
const construct = facts.construct ?? CONSTRUCTED.has(out.document.calleeKind?.(facts))
|
||
const wrapped = continued(node, out.root)
|
||
advance(out, node.startIndex)
|
||
if (wrapped)
|
||
emit(out, { text: '(', role: 'plain', translated: true, raw: '' })
|
||
emit(out, { text: construct ? '新建' : '调用', role: 'storage', translated: true, raw: '' })
|
||
out.gap = true
|
||
out.cursor = facts.callee.startIndex
|
||
const from = out.tokens.length
|
||
collect(facts.callee, span, out)
|
||
for (const token of out.tokens.slice(from)) {
|
||
if (token.offset !== undefined && NAME_ROLES.has(token.role))
|
||
token.role = token.offset !== facts.name?.startIndex ? 'variable' : construct ? 'type' : 'function'
|
||
}
|
||
out.gap = true
|
||
emit(out, { text: `· ${facts.arguments.length} 个参数`, role: 'hint', translated: true, raw: '' })
|
||
if (wrapped)
|
||
emit(out, { text: ')', role: 'plain', translated: true, raw: '' })
|
||
out.cursor = node.endIndex
|
||
}
|
||
|
||
/** 读法用的：照常翻、看节点在不在这一段里、照抄名字、读命令。 */
|
||
function readerContext(out: Collector, span: SourceSpan): ReaderContext {
|
||
const { document } = out
|
||
return {
|
||
document,
|
||
say: node => joinTokens(tokensIn(node, span, document)),
|
||
whole: node => node.startIndex >= span.startIndex && node.endIndex <= span.endIndex,
|
||
name: (node, role) => segment(node.text, role ?? document.language.tokenRole(node), node.startIndex),
|
||
command: (node) => {
|
||
const facts = document.language.command?.(node)
|
||
return facts && commandReading(facts, document)
|
||
},
|
||
}
|
||
}
|
||
|
||
/** 读出来的几段接着前面的词输出；读法给了表、这一段还没排表时表排在这一段下面。 */
|
||
function emitReading(out: Collector, node: Node, reading: Inline[] | { text: Inline[], table?: PreviewTable }): void {
|
||
const { text, table: below } = Array.isArray(reading) ? { text: reading, table: undefined } : reading
|
||
if (below && out.slot && !out.slot.table)
|
||
out.slot.table = below
|
||
emitPieces(out, node, text)
|
||
}
|
||
|
||
function emitPieces(out: Collector, node: Node, pieces: Inline[]): void {
|
||
advance(out, node.startIndex)
|
||
for (const piece of pieces) {
|
||
if (!('role' in piece))
|
||
emit(out, { text: '', role: 'plain', translated: true, raw: '', group: piece })
|
||
else if (piece.text)
|
||
emit(out, { text: piece.text, role: piece.role, translated: true, raw: '', ...(piece.offset !== undefined && { offset: piece.offset }) })
|
||
}
|
||
out.cursor = node.endIndex
|
||
}
|
||
|
||
/**
|
||
* 一条命令：交给命令名的命令规则读（SourceDocument.commandRule）；没有规则、规则不读时命令名照抄（按函数着色、带源码下标）、
|
||
* 各个词照常翻。后面跟着重定向的意思（见 redirectText）；规则给了表（选项写得多的命令）就带着。
|
||
*/
|
||
function commandReading(facts: CommandFacts, document: SourceDocument): CommandReading {
|
||
const say = (node: Node): Inline[] => expression(node, document)
|
||
const ruleFor = (name: CommandWord): CommandRule | undefined => document.commandRule?.(name.literal ?? name.node.text)
|
||
const context: CommandContext = {
|
||
displayLanguage: 'zh',
|
||
translate: say,
|
||
command: (name, args) => ruleFor(name)?.read({ name, arguments: args, redirects: [] }, context),
|
||
}
|
||
const reading = ruleFor(facts.name)?.read(facts, context)
|
||
const name = facts.name.node
|
||
const text = reading?.text ?? [
|
||
segment(facts.name.literal ?? name.text, 'function', name.startIndex),
|
||
...facts.arguments.flatMap(arg => [segment(' ', 'plain', undefined), ...say(arg.node)]),
|
||
]
|
||
return {
|
||
text: [...text, ...facts.redirects.flatMap(redirect => redirectText(redirect, say))],
|
||
...(reading?.value && { value: true }),
|
||
...(reading?.table && { table: reading.table }),
|
||
}
|
||
}
|
||
|
||
function collect(node: Node, span: SourceSpan, out: Collector): void {
|
||
if (node.endIndex <= span.startIndex || node.startIndex >= span.endIndex || node.startIndex === node.endIndex)
|
||
return
|
||
if (node.type === 'comment') {
|
||
emitSource(out, node, { text: '', role: 'comment', translated: false })
|
||
return
|
||
}
|
||
const { language } = out.document
|
||
const iteration = iterationIn(node, span, language)
|
||
if (iteration) {
|
||
emitIteration(out, node, iteration, span)
|
||
return
|
||
}
|
||
const { slot } = out
|
||
const call = slot && expandedCall(node, span, slot, out.document)
|
||
if (slot && call) {
|
||
emitCall(out, slot, node, call, span)
|
||
return
|
||
}
|
||
const reading = table.readers?.[node.type]?.(node, readerContext(out, span))
|
||
if (reading) {
|
||
emitReading(out, node, reading)
|
||
return
|
||
}
|
||
const command = language.command?.(node)
|
||
if (command) {
|
||
emitReading(out, node, commandReading(command, out.document))
|
||
return
|
||
}
|
||
const parts = templateParts(node, span)
|
||
if (parts) {
|
||
advance(out, node.startIndex)
|
||
parts.forEach((part, i) => {
|
||
out.gap ||= i > 0
|
||
if (typeof part === 'string') {
|
||
emit(out, { text: part, role: 'keyword', translated: true, raw: '' })
|
||
return
|
||
}
|
||
out.cursor = part.startIndex
|
||
collect(part, span, out)
|
||
})
|
||
out.cursor = node.endIndex
|
||
return
|
||
}
|
||
const translated = translateNode(node, language)
|
||
if (translated) {
|
||
emitSource(out, node, { ...translated, translated: true })
|
||
return
|
||
}
|
||
if (node.childCount === 0 || stringNodes.has(node.type)) {
|
||
const role = stringNodes.has(node.type) ? 'string' : language.tokenRole(node)
|
||
emitSource(out, node, { text: oneLine(node.text), role, translated: false, offset: offsetOf(node, language) })
|
||
return
|
||
}
|
||
for (const child of node.children) {
|
||
if (child)
|
||
collect(child, span, out)
|
||
}
|
||
}
|
||
|
||
/** 按译文顺序收集 node 落在 span 里的词；给了 slot，跨几行写的调用排成一行、实参摊成表（见 emitCall）。 */
|
||
function tokensIn(node: Node, span: SourceSpan, document: SourceDocument, slot?: TableSlot): Token[] {
|
||
const out: Collector = { document, root: node, tokens: [], cursor: span.startIndex, gap: false, ...(slot && { slot }) }
|
||
collect(node, span, out)
|
||
return out.tokens
|
||
}
|
||
|
||
/** node 的源码原样照抄：叶子按源码着色（给了 role 就都按它），叶子之间的空白、标点照抄。 */
|
||
function verbatim(node: Node, { source, language }: SourceDocument, role?: TokenRole): Segment[] {
|
||
const segments: Segment[] = []
|
||
let cursor = node.startIndex
|
||
const visit = (current: Node): void => {
|
||
if (current.startIndex === current.endIndex)
|
||
return
|
||
if (current.childCount > 0) {
|
||
for (const child of current.children) {
|
||
if (child)
|
||
visit(child)
|
||
}
|
||
return
|
||
}
|
||
pushSegment(segments, source.slice(cursor, current.startIndex), 'plain')
|
||
pushSegment(segments, current.text, role ?? language.tokenRole(current), offsetOf(current, language))
|
||
cursor = current.endIndex
|
||
}
|
||
visit(node)
|
||
pushSegment(segments, source.slice(cursor, node.endIndex), 'plain')
|
||
return segments
|
||
}
|
||
|
||
/** 注释节点去掉注释符号和记号后的原文。 */
|
||
function commentOf(node: Node, { markers }: SourceDocument): string {
|
||
return table.commentText(node, withoutMarkers(node, markers))
|
||
}
|
||
|
||
function translateComment(block: TextBlock, document: SourceDocument): Markdown {
|
||
const { startIndex, endIndex } = block.header
|
||
const root = block.node.tree.rootNode
|
||
const nodes = block.node.type === 'comment' ? commentsWithin(root.descendantForIndex(startIndex, endIndex) ?? root, block.header) : [block.node]
|
||
return { markdown: docText(nodes.map(node => commentOf(node, document))) }
|
||
}
|
||
|
||
function translate(block: TextBlock, document: SourceDocument): TextTranslation {
|
||
if (block.kind === 'comment')
|
||
return { text: translateComment(block, document) }
|
||
const group = block.group && table.groups?.[block.node.type]
|
||
if (block.group && group) {
|
||
const out: Collector = { document, root: block.node, tokens: [], cursor: block.header.startIndex, gap: false }
|
||
return group(block.group, readerContext(out, block.header))
|
||
}
|
||
const slot: TableSlot = {}
|
||
const tokens = tokensIn(block.node, block.header, document, slot)
|
||
const compound = block.children.length > 0 || block.header.endIndex < block.node.endIndex
|
||
const last = tokens.at(-1)
|
||
if (compound && last && !last.translated && HEADER_OPENERS.has(last.raw))
|
||
tokens.pop()
|
||
return { text: joinTokens(tokens), ...(slot.table && { table: slot.table }) }
|
||
}
|
||
|
||
function expression(node: Node, document: SourceDocument): Inline[] {
|
||
return joinTokens(tokensIn(node, nodeSpan(node), document))
|
||
}
|
||
|
||
function doc(node: Node, document: SourceDocument): Markdown | undefined {
|
||
const markdown = docText([commentOf(node, document)])
|
||
return markdown ? { markdown } : undefined
|
||
}
|
||
|
||
function describe(block: FactBlock, document: SourceDocument): PropertyRow[] {
|
||
const { header } = block
|
||
const { language } = document
|
||
// 值里带语句块主体的回调（useMemo(() => { … })）只翻到主体开始之前，和块头部一致
|
||
const clipped = (part: SourceSpan): SourceSpan => part.endIndex > header.endIndex ? { ...part, endIndex: header.endIndex, endRow: header.endRow } : part
|
||
const within = (node: Node, part: SourceSpan): Inline[] => joinTokens(tokensIn(node, clipped(part), document))
|
||
const phrases: Phrases = {
|
||
translated: node => within(node, nodeSpan(node)),
|
||
within,
|
||
expanded: node => expandedText(node, clipped(nodeSpan(node)), document),
|
||
body: node => expression(node, document),
|
||
returned: (node) => {
|
||
const [value] = language.command?.(node)?.arguments ?? []
|
||
return value ? expression(value.node, document) : []
|
||
},
|
||
entryCount: (literal, size) => {
|
||
const name = table.mapLiterals[literal.type]
|
||
return [hint(name ? `${name} ${size} 项` : `${size} 项`)]
|
||
},
|
||
verbatim: node => verbatim(node, document),
|
||
// 点分模块名逐节照抄，每一节各自能跳到定义，和 VS Code 里一样
|
||
module: node => stringNodes.has(node.type) ? [segment(node.text, 'string', offsetOf(node, language))] : verbatim(node, document, 'type'),
|
||
name: (node, role) => [segment(node.text, role, offsetOf(node, language))],
|
||
doc: node => doc(node, document),
|
||
command: (form) => {
|
||
if (form.form === 'exec')
|
||
return expandedText(form.node, clipped(nodeSpan(form.node)), document)
|
||
const reading = document.embedded?.(form.code.node)
|
||
return reading ? embeddedText(reading) : { text: verbatim(form.code.node, document) }
|
||
},
|
||
}
|
||
switch (block.kind) {
|
||
case 'function':
|
||
return functionRows(block.facts, phrases)
|
||
case 'class':
|
||
return classRows(block.facts, phrases)
|
||
case 'enum':
|
||
return enumRows(block.facts, phrases)
|
||
case 'exception':
|
||
return exceptionRows(block.facts, phrases)
|
||
case 'import':
|
||
return importRows(block.facts, phrases)
|
||
case 'interpreter':
|
||
return interpreterRows(block.facts)
|
||
case 'options':
|
||
return optionsRows(block.facts, table.switches ?? {})
|
||
case 'directive':
|
||
return directiveRows(block.facts, table.directives ?? {})
|
||
case 'stage':
|
||
return stageRows(block.facts, phrases)
|
||
case 'health':
|
||
return healthRows(block.facts, phrases)
|
||
case 'start':
|
||
return startRows(block.facts, phrases)
|
||
default:
|
||
return block.facts.scope ? scopedVariableRows(block.facts.declarators, block.facts.scope, phrases) : variableRows(block, phrases)
|
||
}
|
||
}
|
||
|
||
return {
|
||
id: table.id,
|
||
label: table.label,
|
||
languageIds: table.languageIds,
|
||
displayLanguage: 'zh',
|
||
translate,
|
||
describe,
|
||
expression,
|
||
doc,
|
||
attributeNames: ZH_ATTRIBUTE_NAMES,
|
||
}
|
||
}
|
||
|
||
function nodeSpan(node: Node): SourceSpan {
|
||
return { startIndex: node.startIndex, endIndex: node.endIndex, startRow: node.startPosition.row, endRow: node.endPosition.row }
|
||
}
|
||
|
||
function segment(text: string, role: TokenRole, offset: number | undefined): Segment {
|
||
return offset === undefined ? { text, role } : { text, role, offset }
|
||
}
|
||
|
||
/** 相邻的两段不带下标的普通文字并成一段；框不和别的并。 */
|
||
function pushSegment(segments: Inline[], text: string, role: TokenRole, offset?: number): void {
|
||
if (!text)
|
||
return
|
||
const prev = segments.at(-1)
|
||
if (role === 'plain' && prev && 'role' in prev && prev.role === 'plain')
|
||
prev.text += text
|
||
else
|
||
segments.push(segment(text, role, offset))
|
||
}
|
||
|
||
function appendInline(segments: Inline[], pieces: Inline[]): void {
|
||
for (const piece of pieces) {
|
||
if ('role' in piece)
|
||
pushSegment(segments, piece.text, piece.role, piece.offset)
|
||
else
|
||
segments.push(piece)
|
||
}
|
||
}
|
||
|
||
/** 词连成译文：源码里隔着空白的词之间空一格，注释不进译文，右括号前的尾逗号去掉。 */
|
||
function joinTokens(tokens: Token[]): Inline[] {
|
||
const segments: Inline[] = []
|
||
let gap = false
|
||
tokens.forEach((token, i) => {
|
||
gap ||= token.gap
|
||
const trailingComma = token.raw === ',' && !token.translated && CLOSING.has(nextRaw(tokens, i))
|
||
if (isComment(token) || trailingComma) {
|
||
gap ||= token.role === 'comment'
|
||
return
|
||
}
|
||
if (gap && segments.length > 0)
|
||
pushSegment(segments, ' ', 'plain')
|
||
if (token.group)
|
||
segments.push(token.group)
|
||
else
|
||
pushSegment(segments, token.text, token.role, token.offset)
|
||
// 「非」是个词，后面补一个空格，不和操作数粘在一起
|
||
gap = token.translated && token.raw === '!'
|
||
})
|
||
return segments
|
||
}
|
||
|
||
// ─── 属性表 ───────────────────────────────────────────────────────────────────
|
||
|
||
/** 属性表里的值怎么写。 */
|
||
interface Phrases {
|
||
/** 按表翻译，只翻落在块头部里的那一段。 */
|
||
translated: (node: Node) => Inline[]
|
||
/** 同 translated，只翻 node 落在 span 里的那一段（对象字面量里方法的头部）。 */
|
||
within: (node: Node, span: SourceSpan) => Inline[]
|
||
/** 同 translated，里面跨几行写的第一个调用排成一行「新建 / 调用 …」，实参摊成表（见 emitCall）。 */
|
||
expanded: (node: Node) => { text: Inline[], table?: PreviewTable }
|
||
/** 函数主体里的一句（不在块头部里）整句照常翻：函数打印的那几句。 */
|
||
body: (node: Node) => Inline[]
|
||
/** 一句 return 返回的值（语言插件读出的命令的第一个词）；只写了 return 时是空的。 */
|
||
returned: (node: Node) => Inline[]
|
||
/** 字典 / 对象字面量的项数：「字典 3 项」，叫法按字面量的节点类型查表。 */
|
||
entryCount: (literal: Node, size: number) => Segment[]
|
||
/** 源码原样照抄、按源码着色（装饰器）。 */
|
||
verbatim: (node: Node) => Segment[]
|
||
/** 导入来源：字符串字面量按字符串着色，模块名按类型着色。 */
|
||
module: (node: Node) => Segment[]
|
||
/** 源码里的一个名字原样照抄、按 role 着色（函数、类的名称）。 */
|
||
name: (node: Node, role: TokenRole) => Segment[]
|
||
/** 函数、类的说明（JSDoc、docstring）：去掉注释符号、引号和共同缩进的 Markdown 原文；没有文字时没有。 */
|
||
doc: (node: Node) => Markdown | undefined
|
||
/** 一条要执行的命令：数组写法交给命令规则读，shell 写法是宿主读好的嵌着的代码（SourceDocument.embedded），可以带一张选项表。 */
|
||
command: (form: CommandForm) => { text: Inline[], table?: PreviewTable }
|
||
}
|
||
|
||
/**
|
||
* 嵌着的一段代码读成一行：整段只有一步时是这一步的译文（连它的表），几步时各步的译文连起来，
|
||
* 一步失败就停的（shell 的 &&）用「然后」连，别的用分号。
|
||
*/
|
||
export function embeddedText(reading: EmbeddedReading): { text: Inline[], table?: PreviewTable } {
|
||
const [first] = reading.steps
|
||
if (reading.whole && first)
|
||
return first
|
||
const separator: Inline[] = reading.chained ? [segment(' ', 'plain', undefined), segment('然后', 'keyword', undefined), segment(' ', 'plain', undefined)] : [segment('；', 'plain', undefined)]
|
||
return { text: reading.steps.flatMap((step, i) => i > 0 ? [...separator, ...step.text] : step.text) }
|
||
}
|
||
|
||
function hint(text: string): Segment {
|
||
return { text, role: 'hint' }
|
||
}
|
||
|
||
const STREAM_ZH: Record<'output' | 'error' | 'both', string> = { output: '输出', error: '错误输出', both: '全部输出' }
|
||
|
||
/**
|
||
* 重定向的意思，括号里淡色写在命令后面：`>&2` 打印到错误输出，`2>&1` 错误输出并进输出，`2>/dev/null` 丢掉错误输出，
|
||
* `> f` 写到 f，`>> f` 追加到 f，`< f` 从 f 读。
|
||
*/
|
||
function redirectText(redirect: Redirect, say: (node: Node) => Inline[]): Inline[] {
|
||
const note = (...parts: Inline[]): Inline[] => [hint('（'), ...parts, hint('）')]
|
||
switch (redirect.kind) {
|
||
case 'merge':
|
||
if (redirect.stream === redirect.into)
|
||
return []
|
||
return note(hint(redirect.into === 'error' ? '打印到错误输出' : '错误输出并进输出'))
|
||
case 'read':
|
||
return note(hint('从 '), ...say(redirect.source), hint(' 读'))
|
||
default: {
|
||
const stream = STREAM_ZH[redirect.stream]
|
||
if (redirect.target.text === '/dev/null')
|
||
return note(hint(`丢掉${stream}`))
|
||
const where = redirect.kind === 'append' ? '追加到 ' : '写到 '
|
||
return note(hint(redirect.stream === 'output' ? where : `${stream}${where}`), ...say(redirect.target))
|
||
}
|
||
}
|
||
}
|
||
|
||
/** 布尔属性的一行：没有这个概念的语言（facts 里没填）没有这一行。 */
|
||
function flagRows(attribute: 'async' | 'abstract' | 'typeOnly', value: boolean | undefined): PropertyRow[] {
|
||
return value === undefined ? [] : [{ label: ATTRIBUTE_ZH[attribute], value: [{ text: YES_NO[`${value}`], role: 'storage' }] }]
|
||
}
|
||
|
||
function count(n: number): Segment[] {
|
||
return [hint(n > 0 ? `${n} 个` : '无')]
|
||
}
|
||
|
||
function joined(parts: Inline[][], separator: string): Inline[] {
|
||
return parts.flatMap((part, i) => i > 0 ? [{ text: separator, role: 'plain' as const }, ...part] : part)
|
||
}
|
||
|
||
function accessRows(access: Access | undefined): PropertyRow[] {
|
||
return access ? [{ label: ATTRIBUTE_ZH.access, value: [{ text: ACCESS_ZH[access], role: 'keyword' }] }] : []
|
||
}
|
||
|
||
/** 这一行 / 子行讲的定义：几个节点（导入项是名称和别名）合起来在源码里的范围；一个节点都没有就不填。 */
|
||
function defining(...nodes: (Node | undefined)[]): Defines {
|
||
const present = nodes.filter((node): node is Node => node !== undefined)
|
||
if (present.length === 0)
|
||
return {}
|
||
return { defines: { startIndex: Math.min(...present.map(node => node.startIndex)), endIndex: Math.max(...present.map(node => node.endIndex)) } }
|
||
}
|
||
|
||
function importNodes(item: ImportItem): (Node | undefined)[] {
|
||
switch (item.kind) {
|
||
case 'named':
|
||
return [item.name, item.alias]
|
||
case 'module':
|
||
return [item.module, item.alias]
|
||
default:
|
||
return [item.alias]
|
||
}
|
||
}
|
||
|
||
function aliased(item: Inline[], alias: Node | undefined, say: Phrases): Inline[] {
|
||
if (!alias)
|
||
return item
|
||
return [...item, { text: ' ', role: 'plain' }, { text: '作为', role: 'keyword' }, { text: ' ', role: 'plain' }, ...say.translated(alias)]
|
||
}
|
||
|
||
/**
|
||
* 函数：名称 / 说明 / 访问级别 / 异步（只有分这个的语言才有）/ 装饰器 / 输入参数 / 输出参数。
|
||
* 输入参数是一张表，列名是名称 / 类型 / 默认值，每个参数一行；没有默认值的格子留空。
|
||
*/
|
||
function functionRows(facts: FunctionFacts, say: Phrases): PropertyRow[] {
|
||
return [
|
||
{ label: '名称', value: facts.name ? say.name(facts.name, 'function') : [hint('匿名')] },
|
||
...docRows(facts.doc, say),
|
||
...accessRows(facts.access),
|
||
...flagRows('async', facts.async),
|
||
...decoratorRows(facts.decorators, say),
|
||
...(facts.positional
|
||
? positionalRows(facts, facts.positional, say)
|
||
: [
|
||
parameterRow('输入参数', facts.parameters, say),
|
||
{ label: '输出参数', value: facts.returnType ? say.translated(facts.returnType) : [hint('未标注')] },
|
||
]),
|
||
]
|
||
}
|
||
|
||
/**
|
||
* 按位置取参数的函数（shell）：输入参数写「N 个，按位置」，下面一张表「位置 | 函数里叫」（没起名的写 $N）；
|
||
* 输出是它打印的那几句（调用处用 $(函数 …) 拿去用时注明）；返回是写了的 return 值，没写 return、也不打印时是最后一句成不成功。
|
||
*/
|
||
function positionalRows(facts: FunctionFacts, positional: PositionalParameter[], say: Phrases): PropertyRow[] {
|
||
const inputs: PropertyRow = positional.length === 0
|
||
? { label: '输入参数', value: [hint('无')] }
|
||
: {
|
||
label: '输入参数',
|
||
value: [hint(`${positional.length} 个，按位置`)],
|
||
columns: ['位置', '函数里叫'],
|
||
items: positional.map(param => ({
|
||
cells: [[hint(`第 ${param.position} 个`)], param.name ? say.name(param.name, 'variable') : [segment(`$${param.position}`, 'variable', undefined)]],
|
||
...defining(param.name),
|
||
})),
|
||
}
|
||
const prints = facts.prints ?? []
|
||
const captured = facts.captured && facts.name ? [hint(`（调用处用 $(${facts.name.text} …) 拿到）`)] : []
|
||
const output: PropertyRow[] = prints.length > 0 ? [{ label: '输出', value: [...joined(prints.map(say.body), '\n'), ...captured] }] : []
|
||
const returns = facts.returns ?? []
|
||
const values = returns.map(say.returned).filter(value => value.length > 0)
|
||
const result: PropertyRow[] = values.length > 0
|
||
? [{ label: '返回', value: joined(values, '、') }]
|
||
: prints.length === 0 ? [{ label: '返回', value: [hint('最后一句成不成功')] }] : []
|
||
return [inputs, ...output, ...result]
|
||
}
|
||
|
||
/** 解释器：叫什么，怎么找到它（按 PATH 找，或写死的路径）。 */
|
||
function interpreterRows(facts: InterpreterFacts): PropertyRow[] {
|
||
return [
|
||
{ label: '解释器', value: [segment(facts.program, 'function', undefined)] },
|
||
{ label: '怎么找', value: [hint(facts.lookup === 'env' ? `按 PATH 去找（${facts.command}）` : `写死的路径（${facts.command}）`)] },
|
||
]
|
||
}
|
||
|
||
/** 解释器开关：每个开关一行，属性名是开关的意思（没有叫法的写源码里的写法），值是开没开，后面淡色写源码里的写法。 */
|
||
function optionsRows(facts: OptionsFacts, names: Record<string, string>): PropertyRow[] {
|
||
return facts.switches.map(item => ({
|
||
label: names[item.id] ?? item.flag,
|
||
value: [{ text: YES_NO[`${item.on}`], role: 'storage' }, { text: ' ', role: 'plain' }, hint(item.flag)],
|
||
}))
|
||
}
|
||
|
||
/** 解析指令：一条一行，行名是指令的叫法，值照写，后面淡色写补充说明（语法版本是给 BuildKit 看的）。 */
|
||
function directiveRows(facts: DirectiveFacts, names: Record<string, { label: string, note?: string }>): PropertyRow[] {
|
||
return facts.directives.map(({ name, value }) => {
|
||
const known = names[name]
|
||
return { label: known?.label ?? name, value: [segment(value, 'string', undefined), ...(known?.note ? [hint(known.note)] : [])] }
|
||
})
|
||
}
|
||
|
||
/** 选项的值：只是一个变量（`$BUILDPLATFORM`、`${X}`）时按变量着色、能跳到定义，别的照写成字符串。 */
|
||
function flagValue(flag: InstructionFlag): Inline[] {
|
||
const match = /^\$(\{)?(\w+)\}?$/.exec(flag.value)
|
||
if (!match)
|
||
return [segment(flag.value, 'string', flag.valueStart)]
|
||
const name = match[2]
|
||
return [segment(name, /^[A-Z][A-Z0-9_]*$/.test(name) ? 'constant' : 'variable', flag.valueStart + (match[1] ? 2 : 1))]
|
||
}
|
||
|
||
/**
|
||
* 阶段：名称（没起名的写按序号引用）/ 说明（紧贴在 FROM 上面的注释）/ 基础镜像（镜像名，标签、摘要写了才写；
|
||
* 接着前面某个阶段建的写「阶段 X」）/ 平台（写了 --platform 才有）。
|
||
*/
|
||
function stageRows(facts: StageFacts, say: Phrases): PropertyRow[] {
|
||
const doc = (facts.doc ?? []).map(node => say.doc(node)?.markdown).filter(Boolean).join('\n')
|
||
const platform = facts.flags.find(flag => flag.name === 'platform')
|
||
const image: Inline[] = facts.stage
|
||
? [segment('阶段', 'keyword', undefined), segment(' ', 'plain', undefined), ...say.name(facts.image, 'variable')]
|
||
: [
|
||
...say.translated(facts.image),
|
||
...(facts.tag ? [hint('，标签 '), ...say.translated(facts.tag)] : []),
|
||
...(facts.digest ? [hint('，摘要 '), ...say.translated(facts.digest)] : []),
|
||
]
|
||
return [
|
||
facts.name
|
||
? { label: '名称', value: say.name(facts.name, 'variable'), ...defining(facts.name) }
|
||
: { label: '名称', value: [hint(`没起名，按序号 ${facts.index} 引用`)] },
|
||
...(doc ? [{ label: '说明', value: { markdown: doc } }] : []),
|
||
{ label: '基础镜像', value: image },
|
||
...(platform ? [{ label: '平台', value: flagValue(platform) }] : []),
|
||
]
|
||
}
|
||
|
||
const DURATION_UNITS: Record<string, string> = { h: '小时', m: '分', s: '秒', ms: '毫秒', us: '微秒', ns: '纳秒' }
|
||
|
||
/** 时长（`30s`、`1m30s`、`500ms`）：数字按数字着色，单位写成中文；认不出的照写。 */
|
||
function durationText(flag: InstructionFlag): Inline[] {
|
||
const parts = [...flag.value.matchAll(/(\d+(?:\.\d+)?)(ms|us|ns|[hms])/g)]
|
||
if (parts.length === 0 || parts.map(part => part[0]).join('') !== flag.value)
|
||
return [segment(flag.value, 'string', flag.valueStart)]
|
||
return parts.flatMap(([, amount, unit], i) => [...(i > 0 ? [hint(' ')] : []), segment(amount, 'number', undefined), hint(` ${DURATION_UNITS[unit]}`)])
|
||
}
|
||
|
||
/** 一行命令连着它的表（选项写得多的命令的「选项 | 意思」）。 */
|
||
function commandRow(label: string, command: { text: Inline[], table?: PreviewTable }): PropertyRow {
|
||
return { label, value: command.text, ...(command.table && { columns: command.table.columns, items: command.table.items }) }
|
||
}
|
||
|
||
/** 健康检查：间隔 / 超时 / 启动后宽限 / 宽限期里的间隔 / 失败几次算不健康（写了才有）/ 检查命令；NONE 只有一行「不做健康检查」。 */
|
||
function healthRows(facts: HealthFacts, say: Phrases): PropertyRow[] {
|
||
if (facts.none)
|
||
return [{ label: '检查', value: [hint('不做健康检查（基础镜像里设的也关掉）')] }]
|
||
const flag = (name: string): InstructionFlag | undefined => facts.flags.find(item => item.name === name)
|
||
const durations: [string, string][] = [['interval', '间隔'], ['timeout', '超时'], ['start-period', '启动后宽限'], ['start-interval', '宽限期里的间隔']]
|
||
const retries = flag('retries')
|
||
return [
|
||
...durations.flatMap(([name, label]) => {
|
||
const found = flag(name)
|
||
return found ? [{ label, value: durationText(found) }] : []
|
||
}),
|
||
...(retries ? [{ label: '失败几次算不健康', value: [segment(retries.value, 'number', retries.valueStart), hint(' 次')] }] : []),
|
||
...(facts.command ? [commandRow('检查命令', say.command(facts.command))] : []),
|
||
]
|
||
}
|
||
|
||
/** 启动：命令（ENTRYPOINT 叫入口命令）、写法（数组不经过 shell / 经过 shell）。 */
|
||
function startRows(facts: StartFacts, say: Phrases): PropertyRow[] {
|
||
return [
|
||
commandRow(facts.instruction === 'cmd' ? '命令' : '入口命令', say.command(facts.command)),
|
||
{ label: '写法', value: [hint(facts.command.form === 'exec' ? '数组，不经过 shell' : '经过 shell')] },
|
||
]
|
||
}
|
||
|
||
/** 节点所在的那条顶层语句（父节点是根）。 */
|
||
function statementOf(node: Node): Node {
|
||
let current = node
|
||
while (current.parent?.parent)
|
||
current = current.parent
|
||
return current
|
||
}
|
||
|
||
/**
|
||
* 分在哪儿有效的变量（Dockerfile 的构建参数、环境变量）：一个变量是「构建参数 / 环境变量」和「默认值 / 值」两行（构建参数没写默认值的
|
||
* 写「没写，构建时传」）；几个变量（一条写了几个，或几条连着写）并成一行，下面一张「名称 | 值」表，每行讲的定义是那个变量名。
|
||
*/
|
||
function scopedVariableRows(declarators: DeclaratorFacts[], scope: 'build' | 'environment', say: Phrases): PropertyRow[] {
|
||
const label = scope === 'build' ? '构建参数' : '环境变量'
|
||
const valueLabel = scope === 'build' ? '默认值' : '值'
|
||
const valueOf = (declarator: DeclaratorFacts): Inline[] => declarator.value ? say.translated(declarator.value) : [hint(scope === 'build' ? '没写，构建时传' : '空')]
|
||
const [only] = declarators
|
||
if (declarators.length === 1 && only)
|
||
return [{ label, value: say.translated(only.name), ...defining(only.name) }, { label: valueLabel, value: valueOf(only) }]
|
||
const statements = new Set(declarators.map(declarator => statementOf(declarator.name).id)).size
|
||
return [{
|
||
label,
|
||
value: [hint(statements > 1 ? `${statements} 条连着写` : `${declarators.length} 个`)],
|
||
columns: ['名称', valueLabel],
|
||
items: declarators.map(declarator => ({ cells: [say.translated(declarator.name), valueOf(declarator)], ...defining(declarator.name) })),
|
||
}]
|
||
}
|
||
|
||
/** 参数表：列名是名称 / 类型 / 默认值，每个参数一行，讲的定义是参数名；没有默认值的格子留空。 */
|
||
function parameterRow(label: string, parameters: ParameterFacts[], say: Phrases): PropertyRow {
|
||
return {
|
||
label,
|
||
value: count(parameters.length),
|
||
columns: ['名称', '类型', '默认值'],
|
||
items: parameters.map(param => ({
|
||
cells: [
|
||
param.optional ? [...say.translated(param.name), { text: '?', role: 'plain' }] : say.translated(param.name),
|
||
param.type ? say.translated(param.type) : [hint('未标注')],
|
||
param.default ? say.translated(param.default) : [],
|
||
],
|
||
...defining(param.name),
|
||
})),
|
||
}
|
||
}
|
||
|
||
/**
|
||
* 类：名称 / 说明 / 访问级别 / 抽象（只有分这个的语言才有）/ 装饰器 / 类型参数 / 父类 / 实现的接口，
|
||
* 说明及以后各行有才显示；类型参数、父类、实现的接口各是一行，多项用逗号隔开。类型参数那一行讲的是这几个泛型参数的定义。
|
||
*/
|
||
function classRows(facts: ClassFacts, say: Phrases): PropertyRow[] {
|
||
const listRow = (label: string, parts: Inline[][], defines: Defines = {}): PropertyRow[] =>
|
||
parts.length > 0 ? [{ label, value: joined(parts, ', '), ...defines }] : []
|
||
return [
|
||
{ label: '名称', value: facts.name ? say.name(facts.name, 'type') : [hint('匿名')] },
|
||
...docRows(facts.doc, say),
|
||
...accessRows(facts.access),
|
||
...flagRows('abstract', facts.abstract),
|
||
...decoratorRows(facts.decorators, say),
|
||
...listRow('类型参数', facts.typeParameters.map(say.translated), defining(...facts.typeParameters)),
|
||
...listRow('父类', facts.bases.map(base => baseText(base, say))),
|
||
...listRow('实现的接口', facts.implements.map(say.translated)),
|
||
]
|
||
}
|
||
|
||
/** 父类的一项：TypeScript 的泛型实参紧跟在后面。 */
|
||
function baseText(base: BaseFacts, say: Phrases): Inline[] {
|
||
return base.typeArguments ? [...say.translated(base.value), ...say.translated(base.typeArguments)] : say.translated(base.value)
|
||
}
|
||
|
||
/**
|
||
* 枚举（TypeScript 的 enum、Python enum 库的类，同一张表）：名称 / 说明 / 访问级别 / 装饰器 / 继承（照写的父类），有才显示；
|
||
* 成员收成一张表：成员 | 值 | 说明。没写值的成员写语言算出的值（自增的 0、1、2……），算不出来写「自动」，都按说明词着色；
|
||
* 说明是成员的注释、JSDoc、属性 docstring 的原文用空格连起来，一个成员都没有说明时不给这一列。
|
||
*/
|
||
function enumRows(facts: EnumFacts, say: Phrases): PropertyRow[] {
|
||
const docs = facts.members.map(member => docsText(member.docs, say))
|
||
const described = docs.some(Boolean)
|
||
return [
|
||
{ label: '名称', value: say.name(facts.name, 'type') },
|
||
...docRows(facts.doc, say),
|
||
...accessRows(facts.access),
|
||
...decoratorRows(facts.decorators, say),
|
||
...baseRows(facts.bases, say),
|
||
{
|
||
label: '成员',
|
||
value: count(facts.members.length),
|
||
columns: described ? ['成员', '值', '说明'] : ['成员', '值'],
|
||
items: facts.members.map((member, i) => ({
|
||
cells: [
|
||
say.name(member.name, 'variable'),
|
||
member.value ? say.translated(member.value) : [hint(member.computed === undefined ? '自动' : String(member.computed))],
|
||
...(described ? [docCell(docs[i])] : []),
|
||
],
|
||
...defining(member.name),
|
||
})),
|
||
},
|
||
]
|
||
}
|
||
|
||
/**
|
||
* 异常（Python、TypeScript 同一张表）：名称 / 说明 / 访问级别 / 装饰器 / 继承（照写的父类），说明及以后有才显示；
|
||
* 类属性收成一张表：属性 | 类型 | 值 | 说明，一个属性都没写类型时没有「类型」列、都没有说明时没有「说明」列，没写的格子留空；
|
||
* 构造参数同函数的输入参数表。没有类属性、没写构造函数时没有这两行。
|
||
*/
|
||
function exceptionRows(facts: ExceptionFacts, say: Phrases): PropertyRow[] {
|
||
const docs = facts.fields.map(field => docsText(field.docs, say))
|
||
const described = docs.some(Boolean)
|
||
const typed = facts.fields.some(field => field.type)
|
||
const fields: PropertyRow = {
|
||
label: '属性',
|
||
value: count(facts.fields.length),
|
||
columns: ['属性', ...(typed ? ['类型'] : []), '值', ...(described ? ['说明'] : [])],
|
||
items: facts.fields.map((field, i) => ({
|
||
cells: [
|
||
say.name(field.name, 'variable'),
|
||
...(typed ? [field.type ? say.translated(field.type) : []] : []),
|
||
field.value ? say.translated(field.value) : [],
|
||
...(described ? [docCell(docs[i])] : []),
|
||
],
|
||
...defining(field.name),
|
||
})),
|
||
}
|
||
return [
|
||
{ label: '名称', value: facts.name ? say.name(facts.name, 'type') : [hint('匿名')] },
|
||
...docRows(facts.doc, say),
|
||
...accessRows(facts.access),
|
||
...decoratorRows(facts.decorators, say),
|
||
...baseRows(facts.bases, say),
|
||
...(facts.fields.length > 0 ? [fields] : []),
|
||
...(facts.parameters ? [parameterRow('构造参数', facts.parameters, say)] : []),
|
||
]
|
||
}
|
||
|
||
/** 继承：照写的父类，逗号隔开；没有父类时没有这一行。 */
|
||
function baseRows(bases: BaseFacts[], say: Phrases): PropertyRow[] {
|
||
return bases.length > 0 ? [{ label: '继承', value: joined(bases.map(base => baseText(base, say)), ', ') }] : []
|
||
}
|
||
|
||
/** 表格里一格说明：几段说明文字的原文用空格连起来。 */
|
||
function docsText(nodes: Node[], say: Phrases): string {
|
||
return nodes.map(node => say.doc(node)?.markdown).filter(Boolean).join(' ')
|
||
}
|
||
|
||
function docCell(text: string): Inline[] {
|
||
return text ? [segment(text, 'comment', undefined)] : []
|
||
}
|
||
|
||
/** 函数、类、常量、变量的说明：有文字才有这一行。 */
|
||
function docRows(doc: Node | undefined, say: Phrases): PropertyRow[] {
|
||
const markdown = doc && say.doc(doc)
|
||
return markdown ? [{ label: '说明', value: markdown }] : []
|
||
}
|
||
|
||
/** 装饰器：源码原样照抄，多个换行。 */
|
||
function decoratorRows(decorators: Node[], say: Phrases): PropertyRow[] {
|
||
return decorators.length > 0 ? [{ label: '装饰器', value: joined(decorators.map(say.verbatim), '\n') }] : []
|
||
}
|
||
|
||
/** 导入：来源 / 仅类型（只有分这个的语言才有）/ 导入项（每项一行，有别名写成「名称 作为 别名」）。 */
|
||
function importRows(facts: ImportFacts, say: Phrases): PropertyRow[] {
|
||
const itemText = (item: ImportItem): Inline[] => {
|
||
switch (item.kind) {
|
||
case 'named':
|
||
return aliased(say.translated(item.name), item.alias, say)
|
||
case 'default':
|
||
return aliased([hint('默认导出')], item.alias, say)
|
||
case 'namespace':
|
||
return aliased([hint('全部')], item.alias, say)
|
||
case 'module':
|
||
// 一条语句导入多个模块时（Python 的 import a, b）要写出是哪个模块
|
||
return aliased(facts.sources.length > 1 ? say.module(item.module) : [hint('整个模块')], item.alias, say)
|
||
}
|
||
}
|
||
return [
|
||
{ label: '来源', value: facts.sources.length > 0 ? joined(facts.sources.map(say.module), ', ') : [hint('无')] },
|
||
...flagRows('typeOnly', facts.typeOnly),
|
||
{ label: '导入项', value: count(facts.items.length), items: facts.items.map(item => ({ cells: [itemText(item)], ...defining(...importNodes(item)) })) },
|
||
]
|
||
}
|
||
|
||
/**
|
||
* 常量 / 变量：每个声明的变量依次是名称 / 访问级别 / 类型（有标注才有）/ 值（变量叫初始值，见 declaredValue）；
|
||
* 说明（有才显示）是整条语句的，排在第一个名称后面。
|
||
*/
|
||
function variableRows(block: VariableBlock, say: Phrases): PropertyRow[] {
|
||
const valueLabel = block.kind === 'const' ? '值' : '初始值'
|
||
return block.facts.declarators.flatMap((declarator, i) => [
|
||
{ label: '名称', value: say.translated(declarator.name), ...defining(declarator.name) },
|
||
...(i === 0 ? docRows(block.facts.doc, say) : []),
|
||
...accessRows(block.facts.access),
|
||
...(declarator.type ? [{ label: '类型', value: say.translated(declarator.type) }] : []),
|
||
{ label: valueLabel, ...(declarator.value ? declaredValue(declarator.value, declarator.entries, say) : { value: [hint('无')] }) },
|
||
])
|
||
}
|
||
|
||
/**
|
||
* 声明的值：字典 / 对象字面量见 literalValue；别的值是一行译文，里面跨几行写的调用排成「新建 / 调用 …」，
|
||
* 实参摊成表（参数 | 值）展开在这一行下面。
|
||
*/
|
||
function declaredValue(value: Node, entries: EntryFacts[] | undefined, say: Phrases): Pick<PropertyRow, 'value' | 'items' | 'columns'> {
|
||
if (entries?.length)
|
||
return literalValue(value, entries, say)
|
||
const { text, table } = say.expanded(value)
|
||
return table ? { value: text, ...table } : { value: text }
|
||
}
|
||
|
||
/**
|
||
* 值：字典 / 对象字面量（语言插件给了 entries）写项数，各项展开成下一层子行；
|
||
* 空字面量和别的值是一行译文（空的译成「空字典」「空对象」）。
|
||
*/
|
||
function literalValue(value: Node, entries: EntryFacts[] | undefined, say: Phrases): { value: Inline[], items?: PropertyItem[] } {
|
||
if (!entries?.length)
|
||
return { value: say.translated(value) }
|
||
return { value: say.entryCount(value, entries.length), items: entries.map(entry => entryItem(entry, say)) }
|
||
}
|
||
|
||
/**
|
||
* 字典 / 对象字面量的一项是一个子行，两列：键 | 值。键照源码写法（字符串带引号，计算属性名 `[k]` 照常翻），
|
||
* `**x`、`...x` 的键写「展开」、值是展开的表达式，方法的值是它的头部。讲的定义是键（对象的属性名）。
|
||
*/
|
||
function entryItem(entry: EntryFacts, say: Phrases): PropertyItem {
|
||
switch (entry.kind) {
|
||
case 'pair': {
|
||
const { value, ...nested } = literalValue(entry.value, entry.entries, say)
|
||
return { cells: [say.translated(entry.key), value], ...nested, ...defining(entry.key) }
|
||
}
|
||
case 'spread':
|
||
return { cells: [[hint('展开')], say.translated(entry.value)] }
|
||
case 'method':
|
||
return { cells: [say.translated(entry.name), say.within(entry.node, entry.header)], ...defining(entry.name) }
|
||
}
|
||
}
|