Download next.js/packages/next/src/lib/metadata/get-metadata-route.ts from Devendra174/react-code-dataset: direct link, hf CLI and curl.
- Browser
- Download file 4.48 kB
-
https://huggingface.co/datasets/Devendra174/react-code-dataset/resolve/main/next.js/packages/next/src/lib/metadata/get-metadata-route.ts
- Command line
-
hf download hf://datasets/Devendra174/react-code-dataset/next.js/packages/next/src/lib/metadata/get-metadata-route.ts
-
curl -L -o get-metadata-route.ts https://huggingface.co/datasets/Devendra174/react-code-dataset/resolve/main/next.js/packages/next/src/lib/metadata/get-metadata-route.ts
4.48 kB
| import { isMetadataPage } from './is-metadata-route' | |
| import path from '../../shared/lib/isomorphic/path' | |
| import { interpolateDynamicPath } from '../../server/server-utils' | |
| import { getNamedRouteRegex } from '../../shared/lib/router/utils/route-regex' | |
| import { djb2Hash } from '../../shared/lib/hash' | |
| import { normalizeAppPath } from '../../shared/lib/router/utils/app-paths' | |
| import { normalizePathSep } from '../../shared/lib/page-path/normalize-path-sep' | |
| import { | |
| isGroupSegment, | |
| isParallelRouteSegment, | |
| } from '../../shared/lib/segment' | |
| /* | |
| * If there's special convention like (...) or @ in the page path, | |
| * Give it a unique hash suffix to avoid conflicts | |
| * | |
| * e.g. | |
| * /opengraph-image -> /opengraph-image | |
| * /(post)/opengraph-image.tsx -> /opengraph-image-[0-9a-z]{6} | |
| * | |
| * Sitemap is an exception, it should not have a suffix. | |
| * Each sitemap contains all the urls of sub routes, we don't have the case of duplicates `/(group)/sitemap.[ext]` and `/sitemap.[ext]` since they should be the same. | |
| * Hence we always normalize the urls for sitemap and do not append hash suffix, and ensure user-land only contains one sitemap per pathname. | |
| * | |
| * /sitemap -> /sitemap | |
| * /(post)/sitemap -> /sitemap | |
| */ | |
| function getMetadataRouteSuffix(page: string) { | |
| // Remove the last segment and get the parent pathname | |
| // e.g. /parent/a/b/c -> /parent/a/b | |
| // e.g. /parent/opengraph-image -> /parent | |
| const parentPathname = path.dirname(page) | |
| // Only apply suffix to metadata routes except for sitemaps | |
| if (page.endsWith('/sitemap')) { | |
| return '' | |
| } | |
| // Calculate the hash suffix based on the parent path | |
| let suffix = '' | |
| // Check if there's any special characters in the parent pathname. | |
| const segments = parentPathname.split('/') | |
| if ( | |
| segments.some((seg) => isGroupSegment(seg) || isParallelRouteSegment(seg)) | |
| ) { | |
| // Hash the parent path to get a unique suffix | |
| suffix = djb2Hash(parentPathname).toString(36).slice(0, 6) | |
| } | |
| return suffix | |
| } | |
| /** | |
| * Fill the dynamic segment in the metadata route | |
| * | |
| * Example: | |
| * fillMetadataSegment('/a/[slug]', { params: { slug: 'b' } }, 'open-graph') -> '/a/b/open-graph' | |
| * | |
| */ | |
| export function fillMetadataSegment( | |
| segment: string, | |
| params: any, | |
| lastSegment: string | |
| ) { | |
| const pathname = normalizeAppPath(segment) | |
| const routeRegex = getNamedRouteRegex(pathname, { | |
| prefixRouteKeys: false, | |
| }) | |
| const route = interpolateDynamicPath(pathname, params, routeRegex) | |
| const { name, ext } = path.parse(lastSegment) | |
| const pagePath = path.posix.join(segment, name) | |
| const suffix = getMetadataRouteSuffix(pagePath) | |
| const routeSuffix = suffix ? `-${suffix}` : '' | |
| return normalizePathSep(path.join(route, `${name}${routeSuffix}${ext}`)) | |
| } | |
| /** | |
| * Map metadata page key to the corresponding route | |
| * | |
| * static file page key: /app/robots.txt -> /robots.xml -> /robots.txt/route | |
| * dynamic route page key: /app/robots.tsx -> /robots -> /robots.txt/route | |
| * | |
| * @param page | |
| * @returns | |
| */ | |
| export function normalizeMetadataRoute(page: string) { | |
| if (!isMetadataPage(page)) { | |
| return page | |
| } | |
| let route = page | |
| let suffix = '' | |
| if (page === '/robots') { | |
| route += '.txt' | |
| } else if (page === '/manifest') { | |
| route += '.webmanifest' | |
| } else { | |
| suffix = getMetadataRouteSuffix(page) | |
| } | |
| // Support both /<metadata-route.ext> and custom routes /<metadata-route>/route.ts. | |
| // If it's a metadata file route, we need to append /[id]/route to the page. | |
| if (!route.endsWith('/route')) { | |
| const { dir, name: baseName, ext } = path.parse(route) | |
| route = path.posix.join( | |
| dir, | |
| `${baseName}${suffix ? `-${suffix}` : ''}${ext}`, | |
| 'route' | |
| ) | |
| } | |
| return route | |
| } | |
| // Normalize metadata route page to either a single route or a dynamic route. | |
| // e.g. Input: /sitemap/route | |
| // when isDynamic is false, single route -> /sitemap.xml/route | |
| // when isDynamic is false, dynamic route -> /sitemap/[__metadata_id__]/route | |
| // also works for pathname such as /sitemap -> /sitemap.xml, but will not append /route suffix | |
| export function normalizeMetadataPageToRoute(page: string, isDynamic: boolean) { | |
| const isRoute = page.endsWith('/route') | |
| const routePagePath = isRoute ? page.slice(0, -'/route'.length) : page | |
| const metadataRouteExtension = routePagePath.endsWith('/sitemap') | |
| ? '.xml' | |
| : '' | |
| const mapped = isDynamic | |
| ? `${routePagePath}/[__metadata_id__]` | |
| : `${routePagePath}${metadataRouteExtension}` | |
| return mapped + (isRoute ? '/route' : '') | |
| } | |