UNPKG

langium

Version:

A language engineering tool for the Language Server Protocol

271 lines (244 loc) 13.4 kB
/****************************************************************************** * Copyright 2021 TypeFox GmbH * This program and the accompanying materials are made available under the * terms of the MIT License, which is available in the project root. ******************************************************************************/ import type { CodeDescription, DiagnosticRelatedInformation, DiagnosticTag, integer, Range } from 'vscode-languageserver-types'; import { assertUnreachable } from '../index.js'; import type { LangiumCoreServices } from '../services.js'; import type { AstNode, AstReflection, Properties } from '../syntax-tree.js'; import type { CancellationToken } from '../utils/cancellation.js'; import { MultiMap } from '../utils/collections.js'; import type { MaybePromise } from '../utils/promise-utils.js'; import { isOperationCancelled } from '../utils/promise-utils.js'; import type { Stream } from '../utils/stream.js'; import { stream } from '../utils/stream.js'; import type { DocumentSegment } from '../workspace/documents.js'; export type DiagnosticInfo<N extends AstNode, P extends string = Properties<N>> = { /** The AST node to which the diagnostic is attached. */ node: N; /** If a property name is given, the diagnostic is restricted to the corresponding text region. */ property?: P; /** If the value of a keyword is given, the diagnostic will appear at its corresponding text region */ keyword?: string; /** In case of a multi-value property (array), an index can be given to select a specific element. */ index?: number; /** If you want to create a diagnostic independent to any property, use the range property. */ range?: Range; /** The diagnostic's code, which usually appear in the user interface. */ code?: integer | string; /** An optional property to describe the error code. */ codeDescription?: CodeDescription; /** Additional metadata about the diagnostic. */ tags?: DiagnosticTag[]; /** An array of related diagnostic information, e.g. when symbol-names within a scope collide all definitions can be marked via this property. */ relatedInformation?: DiagnosticRelatedInformation[]; /** A data entry field that is preserved between a `textDocument/publishDiagnostics` notification and `textDocument/codeAction` request. */ data?: unknown; } /** * Shape of information commonly used in the `data` field of diagnostics. */ export interface DiagnosticData { /** Diagnostic code for identifying which code action to apply. This code is _not_ shown in the user interface. */ code: string /** Specifies where to apply the code action in the form of a `DocumentSegment`. */ actionSegment?: DocumentSegment /** Specifies where to apply the code action in the form of a `Range`. */ actionRange?: Range } /** * Create DiagnosticData for a given diagnostic code. The result can be put into the `data` field of a DiagnosticInfo. */ export function diagnosticData(code: string): DiagnosticData { return { code }; } export type ValidationSeverity = 'error' | 'warning' | 'info' | 'hint'; export type ValidationAcceptor = <N extends AstNode>(severity: ValidationSeverity, message: string, info: DiagnosticInfo<N>) => void export type ValidationCheck<T extends AstNode = AstNode> = (node: T, accept: ValidationAcceptor, cancelToken: CancellationToken) => MaybePromise<void>; /** * A utility type for describing functions which will be called once before or after all the AstNodes of an AST/Langium document are validated. * * The AST is represented by its root AstNode. * * The given validation acceptor helps to report some early or lately detected issues. * * The 'categories' indicate, which validation categories are executed for all the AstNodes. * This helps to tailor the preparations/tear-down logic to the actually executed checks on the nodes. * * It is recommended to support interrupts during long-running logic with 'interruptAndCheck(cancelToken)'. */ export type ValidationPreparation = (rootNode: AstNode, accept: ValidationAcceptor, categories: ValidationCategory[], cancelToken: CancellationToken) => MaybePromise<void>; /** * A utility type for associating non-primitive AST types to corresponding validation checks. For example: * * ```ts * const checks: ValidationChecks<StatemachineAstType> = { * State: validator.checkStateNameStartsWithCapital * }; * ``` * * If an AST type does not extend AstNode, e.g. if it describes a union of string literals, that type's name must not occur as a key in objects of type `ValidationCheck<...>`. * * @param T a type definition mapping language specific type names (keys) to the corresponding types (values) */ export type ValidationChecks<T> = { [K in keyof T]?: T[K] extends AstNode ? ValidationCheck<T[K]> | Array<ValidationCheck<T[K]>> : never } & { AstNode?: ValidationCheck<AstNode> | Array<ValidationCheck<AstNode>>; } /** * There are 3 pre-defined categories: `fast`, `slow` and `built-in`. * * `fast` checks can be executed after every document change (i.e. as the user is typing). If a check * is too slow it can delay the response to document changes, yielding bad user experience. By marking * it as `slow`, it will be skipped for normal as-you-type validation. Then it's up to you when to * schedule these long-running checks: after the fast checks are done, or after saving a document, * or with an explicit command, etc. * * `built-in` checks are errors produced by the lexer, the parser, or the linker. They cannot be used * for custom validation checks. * * You can also provide user-defined categories. These check will be skipped by default. Then it's up * to you to schedule these checks: after the fast checks are done, or after saving a document, * or with an explicit command, etc. */ export type ValidationCategory = 'fast' | 'slow' | 'built-in' | (string & {}); export namespace ValidationCategory { export const all: readonly ValidationCategory[] = ['fast', 'slow', 'built-in']; } type ValidationCheckEntry = { check: ValidationCheck category: ValidationCategory } /** * Manages a set of `ValidationCheck`s to be applied when documents are validated. */ export class ValidationRegistry { private readonly entries = new MultiMap<string, ValidationCheckEntry>(); private readonly reflection: AstReflection; private entriesBefore: ValidationPreparation[] = []; private entriesAfter: ValidationPreparation[] = []; constructor(services: LangiumCoreServices) { this.reflection = services.shared.AstReflection; } /** * Register a set of validation checks. Each value in the record can be either a single validation check (i.e. a function) * or an array of validation checks. * * @param checksRecord Set of validation checks to register. * @param category Optional category for the validation checks (defaults to `'fast'`). * @param thisObj Optional object to be used as `this` when calling the validation check functions. */ register<T>(checksRecord: ValidationChecks<T>, thisObj: ThisParameterType<unknown> = this, category: ValidationCategory = 'fast'): void { if (category === 'built-in') { throw new Error("The 'built-in' category is reserved for lexer, parser, and linker errors."); } for (const [type, ch] of Object.entries(checksRecord)) { const callbacks = ch as ValidationCheck | ValidationCheck[]; if (Array.isArray(callbacks)) { for (const check of callbacks) { const entry: ValidationCheckEntry = { check: this.wrapValidationException(check, thisObj), category }; this.addEntry(type, entry); } } else if (typeof callbacks === 'function') { const entry: ValidationCheckEntry = { check: this.wrapValidationException(callbacks, thisObj), category }; this.addEntry(type, entry); } else { assertUnreachable(callbacks); } } } protected wrapValidationException(check: ValidationCheck, thisObj: unknown): ValidationCheck { return async (node, accept, cancelToken) => { await this.handleException(() => check.call(thisObj, node, accept, cancelToken), 'An error occurred during validation', accept, node); }; } protected async handleException(functionality: () => MaybePromise<void>, messageContext: string, accept: ValidationAcceptor, node: AstNode): Promise<void> { try { await functionality(); } catch (err) { if (isOperationCancelled(err)) { throw err; } console.error(`${messageContext}:`, err); if (err instanceof Error && err.stack) { console.error(err.stack); } const messageDetails = err instanceof Error ? err.message : String(err); accept('error', `${messageContext}: ${messageDetails}`, { node }); } } protected addEntry(type: string, entry: ValidationCheckEntry): void { if (type === 'AstNode') { this.entries.add('AstNode', entry); return; } for (const subtype of this.reflection.getAllSubTypes(type)) { this.entries.add(subtype, entry); } } getChecks(type: string, categories?: ValidationCategory[]): Stream<ValidationCheck> { let checks = stream(this.entries.get(type)) .concat(this.entries.get('AstNode')); if (categories) { checks = checks.filter(entry => categories.includes(entry.category)); } return checks.map(entry => entry.check); } /** * Register logic which will be executed once before validating all the nodes of an AST/Langium document. * This helps to prepare or initialize some information which are required or reusable for the following checks on the AstNodes. * * As an example, for validating unique fully-qualified names of nodes in the AST, * here the map for mapping names to nodes could be established. * During the usual checks on the nodes, they are put into this map with their name. * * Note that this approach makes validations stateful, which is relevant e.g. when cancelling the validation. * Therefore it is recommended to clear stored information * _before_ validating an AST to validate each AST unaffected from other ASTs * AND _after_ validating the AST to free memory by information which are no longer used. * * @param checkBefore a set-up function which will be called once before actually validating an AST * @param thisObj Optional object to be used as `this` when calling the validation check functions. */ registerBeforeDocument(checkBefore: ValidationPreparation, thisObj: ThisParameterType<unknown> = this): void { this.entriesBefore.push(this.wrapPreparationException(checkBefore, 'An error occurred during set-up of the validation', thisObj)); } /** * Register logic which will be executed once after validating all the nodes of an AST/Langium document. * This helps to finally evaluate information which are collected during the checks on the AstNodes. * * As an example, for validating unique fully-qualified names of nodes in the AST, * here the map with all the collected nodes and their names is checked * and validation hints are created for all nodes with the same name. * * Note that this approach makes validations stateful, which is relevant e.g. when cancelling the validation. * Therefore it is recommended to clear stored information * _before_ validating an AST to validate each AST unaffected from other ASTs * AND _after_ validating the AST to free memory by information which are no longer used. * * @param checkBefore a set-up function which will be called once before actually validating an AST * @param thisObj Optional object to be used as `this` when calling the validation check functions. */ registerAfterDocument(checkAfter: ValidationPreparation, thisObj: ThisParameterType<unknown> = this): void { this.entriesAfter.push(this.wrapPreparationException(checkAfter, 'An error occurred during tear-down of the validation', thisObj)); } protected wrapPreparationException(check: ValidationPreparation, messageContext: string, thisObj: unknown): ValidationPreparation { return async (rootNode, accept, categories, cancelToken) => { await this.handleException(() => check.call(thisObj, rootNode, accept, categories, cancelToken), messageContext, accept, rootNode); }; } get checksBefore(): ValidationPreparation[] { return this.entriesBefore; } get checksAfter(): ValidationPreparation[] { return this.entriesAfter; } }