| 1 | /** |
| 2 | * Copyright (c) Meta Platforms, Inc. and affiliates. |
| 3 | * |
| 4 | * This source code is licensed under the MIT license found in the |
| 5 | * LICENSE file in the root directory of this source tree. |
| 6 | */ |
| 7 | |
| 8 | import type { |
| 9 | ChangedFile, |
| 10 | CodeReviewSystem, |
| 11 | CommitInfo, |
| 12 | CommitPhaseType, |
| 13 | Hash, |
| 14 | RepoRelativePath, |
| 15 | ShelvedChange, |
| 16 | SmartlogCommits, |
| 17 | StableCommitFetchConfig, |
| 18 | StableInfo, |
| 19 | SuccessorInfo, |
| 20 | } from 'isl/src/types'; |
| 21 | import type {Logger} from './logger'; |
| 22 | |
| 23 | import path from 'path'; |
| 24 | import {Internal} from './Internal'; |
| 25 | import {MAX_FETCHED_FILES_PER_COMMIT} from './commands'; |
| 26 | import {fromEntries} from './utils'; |
| 27 | |
| 28 | export const COMMIT_END_MARK = '<<COMMIT_END_MARK>>'; |
| 29 | export const NULL_CHAR = '\0'; |
| 30 | export const ESCAPED_NULL_CHAR = '\\0'; |
| 31 | export const WDIR_PARENT_MARKER = '@'; |
| 32 | |
| 33 | ///// Main commits fetch ///// |
| 34 | |
| 35 | export const mainFetchTemplateFields = (codeReviewSystem: CodeReviewSystem) => ({ |
| 36 | hash: '{node}', |
| 37 | title: '{desc|firstline}', |
| 38 | author: '{author}', |
| 39 | // We prefer committerdate over authordate as authordate sometimes makes |
| 40 | // amended or rebased commits look stale |
| 41 | date: '{committerdate|isodatesec}', |
| 42 | phase: '{phase}', |
| 43 | bookmarks: `{bookmarks % '{bookmark}${ESCAPED_NULL_CHAR}'}`, |
| 44 | remoteBookmarks: `{remotenames % '{remotename}${ESCAPED_NULL_CHAR}'}`, |
| 45 | parents: `{parents % "{node}${ESCAPED_NULL_CHAR}"}`, |
| 46 | grandparents: `{grandparents % "{node}${ESCAPED_NULL_CHAR}"}`, |
| 47 | isDot: `{ifcontains(rev, revset('.'), '${WDIR_PARENT_MARKER}')}`, |
| 48 | // We don't need files for public commits, and public commits are sometimes gigantic codemods without you realizing. |
| 49 | // No need to fetch if not draft. |
| 50 | files: `{ifeq(phase, 'draft', join(files,'${ESCAPED_NULL_CHAR}'), '')}`, |
| 51 | totalFileCount: '{files|count}', // We skip getting files for public commits, but we still want to know how many files there would be |
| 52 | successorInfo: '{mutations % "{operation}:{successors % "{node}"},"}', |
| 53 | closestPredecessors: '{predecessors % "{node},"}', |
| 54 | // This would be more elegant as a new built-in template |
| 55 | diffId: |
| 56 | codeReviewSystem.type === 'phabricator' |
| 57 | ? '{phabdiff}' |
| 58 | : codeReviewSystem.type === 'github' |
| 59 | ? '{github_pull_request_number}' |
| 60 | : '', |
| 61 | isFollower: '{sapling_pr_follower|json}', |
| 62 | stableCommitMetadata: Internal.stableCommitConfig?.template ?? '', |
| 63 | // Description must be last |
| 64 | description: '{desc}', |
| 65 | }); |
| 66 | |
| 67 | export function getMainFetchTemplate(codeReviewSystem: CodeReviewSystem): string { |
| 68 | return [...Object.values(mainFetchTemplateFields(codeReviewSystem)), COMMIT_END_MARK].join('\n'); |
| 69 | } |
| 70 | |
| 71 | /** |
| 72 | * Extract CommitInfos from log calls that use FETCH_TEMPLATE. |
| 73 | */ |
| 74 | export function parseCommitInfoOutput( |
| 75 | logger: Logger, |
| 76 | output: string, |
| 77 | reviewSystem: CodeReviewSystem, |
| 78 | stableCommitConfig = Internal.stableCommitConfig as StableCommitFetchConfig | null, |
| 79 | ): SmartlogCommits { |
| 80 | const fields = mainFetchTemplateFields(reviewSystem); |
| 81 | const index = fromEntries(Object.keys(fields).map((key, i) => [key, i])) as { |
| 82 | [key in Required<keyof typeof fields>]: number; |
| 83 | }; |
| 84 | |
| 85 | const revisions = output.split(COMMIT_END_MARK); |
| 86 | const commitInfos: Array<CommitInfo> = []; |
| 87 | for (const chunk of revisions) { |
| 88 | try { |
| 89 | const lines = chunk.trimStart().split('\n'); |
| 90 | if (lines.length < Object.keys(fields).length) { |
| 91 | continue; |
| 92 | } |
| 93 | const files = lines[index.files].split(NULL_CHAR).filter(e => e.length > 0); |
| 94 | |
| 95 | // Find if the commit is entirely within the cwd and therefore more relevant to the user. |
| 96 | // Note: this must be done on the server using the full list of files, not just the sample that the client gets. |
| 97 | // TODO: should we cache this by commit hash to avoid iterating all files on the same commits every time? |
| 98 | const maxCommonPathPrefix = findMaxCommonPathPrefix(files); |
| 99 | |
| 100 | commitInfos.push({ |
| 101 | hash: lines[index.hash], |
| 102 | title: lines[index.title], |
| 103 | author: lines[index.author], |
| 104 | date: new Date(lines[index.date]), |
| 105 | parents: splitLine(lines[index.parents]) as string[], |
| 106 | grandparents: splitLine(lines[index.grandparents]) as string[], |
| 107 | phase: lines[index.phase] as CommitPhaseType, |
| 108 | bookmarks: splitLine(lines[index.bookmarks]), |
| 109 | remoteBookmarks: splitLine(lines[index.remoteBookmarks]), |
| 110 | isDot: lines[index.isDot] === WDIR_PARENT_MARKER, |
| 111 | filePathsSample: files.slice(0, MAX_FETCHED_FILES_PER_COMMIT), |
| 112 | totalFileCount: parseInt(lines[index.totalFileCount], 10), |
| 113 | successorInfo: parseSuccessorData(lines[index.successorInfo]), |
| 114 | closestPredecessors: splitLine(lines[index.closestPredecessors], ','), |
| 115 | description: lines |
| 116 | .slice(index.description + 1 /* first field of description is title; skip it */) |
| 117 | .join('\n') |
| 118 | .trim(), |
| 119 | diffId: lines[index.diffId] != '' ? lines[index.diffId] : undefined, |
| 120 | isFollower: lines[index.isFollower] !== '' ? (JSON.parse(lines[index.isFollower]) as boolean) : false, |
| 121 | stableCommitMetadata: |
| 122 | lines[index.stableCommitMetadata] != '' |
| 123 | ? stableCommitConfig?.parse(lines[index.stableCommitMetadata]) |
| 124 | : undefined, |
| 125 | maxCommonPathPrefix, |
| 126 | }); |
| 127 | } catch (err) { |
| 128 | logger.error('failed to parse commit', err); |
| 129 | } |
| 130 | } |
| 131 | return commitInfos; |
| 132 | } |
| 133 | |
| 134 | /** |
| 135 | * Given a set of changed files, find the longest common path prefix. |
| 136 | * See {@link CommitInfo}.maxCommonPathPrefix |
| 137 | * TODO: This could be cached by commit hash |
| 138 | */ |
| 139 | export function findMaxCommonPathPrefix(filePaths: Array<RepoRelativePath>): RepoRelativePath { |
| 140 | let max: null | Array<string> = null; |
| 141 | let maxLength = 0; |
| 142 | |
| 143 | // Path module separator should match what `sl` gives us |
| 144 | const sep = path.sep; |
| 145 | |
| 146 | for (const path of filePaths) { |
| 147 | if (max == null) { |
| 148 | max = path.split(sep); |
| 149 | max.pop(); // ignore file part, only care about directory |
| 150 | maxLength = max.reduce((acc, part) => acc + part.length + 1, 0); // +1 for slash |
| 151 | continue; |
| 152 | } |
| 153 | // small optimization: we only need to look as long as the max so far, max common path will always be shorter |
| 154 | const parts = path.slice(0, maxLength).split(sep); |
| 155 | for (const [i, part] of parts.entries()) { |
| 156 | if (part !== max[i]) { |
| 157 | max = max.slice(0, i); |
| 158 | maxLength = max.reduce((acc, part) => acc + part.length + 1, 0); // +1 for slash |
| 159 | break; |
| 160 | } |
| 161 | } |
| 162 | if (max.length === 0) { |
| 163 | return ''; // we'll never get *more* specific, early exit |
| 164 | } |
| 165 | } |
| 166 | |
| 167 | const result = (max ?? []).join(sep); |
| 168 | if (result == '') { |
| 169 | return result; |
| 170 | } |
| 171 | return result + sep; |
| 172 | } |
| 173 | |
| 174 | /** |
| 175 | * Additional stable locations in the commit fetch will not automatically |
| 176 | * include "stableCommitMetadata". Insert this data onto the commits. |
| 177 | */ |
| 178 | export function attachStableLocations(commits: Array<CommitInfo>, locations: Array<StableInfo>) { |
| 179 | const map: Record<Hash, Array<StableInfo>> = {}; |
| 180 | for (const location of locations) { |
| 181 | const existing = map[location.hash] ?? []; |
| 182 | map[location.hash] = [...existing, location]; |
| 183 | } |
| 184 | |
| 185 | for (const commit of commits) { |
| 186 | if (commit.hash in map) { |
| 187 | commit.stableCommitMetadata = [ |
| 188 | ...(commit.stableCommitMetadata ?? []), |
| 189 | ...map[commit.hash].map(location => ({ |
| 190 | value: location.name, |
| 191 | description: location.info ?? '', |
| 192 | })), |
| 193 | ]; |
| 194 | } |
| 195 | } |
| 196 | } |
| 197 | |
| 198 | ///// Shelve ///// |
| 199 | |
| 200 | export const SHELVE_FIELDS = { |
| 201 | hash: '{node}', |
| 202 | name: '{shelvename}', |
| 203 | author: '{author}', |
| 204 | date: '{date|isodatesec}', |
| 205 | filesAdded: '{file_adds|json}', |
| 206 | filesModified: '{file_mods|json}', |
| 207 | filesRemoved: '{file_dels|json}', |
| 208 | description: '{desc}', |
| 209 | }; |
| 210 | export const SHELVE_FIELD_INDEX = fromEntries( |
| 211 | Object.keys(SHELVE_FIELDS).map((key, i) => [key, i]), |
| 212 | ) as { |
| 213 | [key in Required<keyof typeof SHELVE_FIELDS>]: number; |
| 214 | }; |
| 215 | export const SHELVE_FETCH_TEMPLATE = [...Object.values(SHELVE_FIELDS), COMMIT_END_MARK].join('\n'); |
| 216 | |
| 217 | export function parseShelvedCommitsOutput(logger: Logger, output: string): Array<ShelvedChange> { |
| 218 | const shelves = output.split(COMMIT_END_MARK); |
| 219 | const commitInfos: Array<ShelvedChange> = []; |
| 220 | for (const chunk of shelves) { |
| 221 | try { |
| 222 | const lines = chunk.trim().split('\n'); |
| 223 | if (lines.length < Object.keys(SHELVE_FIELDS).length) { |
| 224 | continue; |
| 225 | } |
| 226 | const files: Array<ChangedFile> = [ |
| 227 | ...(JSON.parse(lines[SHELVE_FIELD_INDEX.filesModified]) as Array<string>).map(path => ({ |
| 228 | path, |
| 229 | status: 'M' as const, |
| 230 | })), |
| 231 | ...(JSON.parse(lines[SHELVE_FIELD_INDEX.filesAdded]) as Array<string>).map(path => ({ |
| 232 | path, |
| 233 | status: 'A' as const, |
| 234 | })), |
| 235 | ...(JSON.parse(lines[SHELVE_FIELD_INDEX.filesRemoved]) as Array<string>).map(path => ({ |
| 236 | path, |
| 237 | status: 'R' as const, |
| 238 | })), |
| 239 | ]; |
| 240 | commitInfos.push({ |
| 241 | hash: lines[SHELVE_FIELD_INDEX.hash], |
| 242 | name: lines[SHELVE_FIELD_INDEX.name], |
| 243 | date: new Date(lines[SHELVE_FIELD_INDEX.date]), |
| 244 | filesSample: files.slice(0, MAX_FETCHED_FILES_PER_COMMIT), |
| 245 | totalFileCount: files.length, |
| 246 | description: lines.slice(SHELVE_FIELD_INDEX.description).join('\n'), |
| 247 | }); |
| 248 | } catch (err) { |
| 249 | logger.error('failed to parse shelved change'); |
| 250 | } |
| 251 | } |
| 252 | return commitInfos; |
| 253 | } |
| 254 | |
| 255 | ///// Changed Files ///// |
| 256 | |
| 257 | export const CHANGED_FILES_FIELDS = { |
| 258 | hash: '{node}', |
| 259 | filesAdded: '{file_adds|json}', |
| 260 | filesModified: '{file_mods|json}', |
| 261 | filesRemoved: '{file_dels|json}', |
| 262 | }; |
| 263 | export const CHANGED_FILES_INDEX = fromEntries( |
| 264 | Object.keys(CHANGED_FILES_FIELDS).map((key, i) => [key, i]), |
| 265 | ) as { |
| 266 | [key in Required<keyof typeof CHANGED_FILES_FIELDS>]: number; |
| 267 | }; |
| 268 | export const CHANGED_FILES_TEMPLATE = [ |
| 269 | ...Object.values(CHANGED_FILES_FIELDS), |
| 270 | COMMIT_END_MARK, |
| 271 | ].join('\n'); |
| 272 | |
| 273 | ///// Helpers ///// |
| 274 | |
| 275 | function parseSuccessorData(successorData: string): SuccessorInfo | undefined { |
| 276 | const [successorString] = successorData.split(',', 1); // we're only interested in the first available mutation |
| 277 | if (!successorString) { |
| 278 | return undefined; |
| 279 | } |
| 280 | const successor = successorString.split(':'); |
| 281 | return { |
| 282 | hash: successor[1], |
| 283 | type: successor[0], |
| 284 | }; |
| 285 | } |
| 286 | function splitLine(line: string, separator = NULL_CHAR): Array<string> { |
| 287 | return line.split(separator).filter(e => e.length > 0); |
| 288 | } |
| 289 | |