git.stevedylan.dev

packages/cli/src/commands/inject.ts 6.1 K raw

import * as fs from "fs/promises";
import { command, flag, option, optional, string } from "cmd-ts";
import { log } from "@clack/prompts";
import * as path from "path";
import { glob } from "glob";
import { loadConfig, loadState, findConfig } from "../lib/config";

export const injectCommand = command({
	name: "inject",
	description:
		"Inject site.standard.document link tags into built HTML files",
	args: {
		outputDir: option({
			long: "output",
			short: "o",
			description: "Output directory to scan for HTML files",
			type: optional(string),
		}),
		dryRun: flag({
			long: "dry-run",
			short: "n",
			description: "Preview what would be injected without making changes",
		}),
	},
	handler: async ({ outputDir: outputDirArg, dryRun }) => {
		// Load config
		const configPath = await findConfig();
		if (!configPath) {
			log.error("No sequoia.json found. Run 'sequoia init' first.");
			process.exit(1);
		}

		const config = await loadConfig(configPath);
		const configDir = path.dirname(configPath);

		// Determine output directory
		const outputDir = outputDirArg || config.outputDir || "./dist";
		const resolvedOutputDir = path.isAbsolute(outputDir)
			? outputDir
			: path.join(configDir, outputDir);

		log.info(`Scanning for HTML files in: ${resolvedOutputDir}`);

		// Load state to get atUri mappings
		const state = await loadState(configDir);

		// Generic filenames where the slug is the parent directory, not the filename
		// Covers: SvelteKit (+page), Astro/Hugo (index), Next.js (page), etc.
		const genericFilenames = new Set([
			"+page",
			"index",
			"_index",
			"page",
			"readme",
		]);

		// Build a map of slug/path to atUri from state
		const pathToAtUri = new Map<string, string>();
		for (const [filePath, postState] of Object.entries(state.posts)) {
			if (postState.atUri) {
				// Extract slug from file path (e.g., ./content/blog/my-post.md -> my-post)
				let basename = path.basename(filePath, path.extname(filePath));

				// If the filename is a generic convention name, use the parent directory as slug
				if (genericFilenames.has(basename.toLowerCase())) {
					// Split path and filter out route groups like (blog-article)
					const pathParts = filePath
						.split(/[/\\]/)
						.filter((p) => p && !(p.startsWith("(") && p.endsWith(")")));
					// The slug should be the second-to-last part (last is the filename)
					if (pathParts.length >= 2) {
						const slug = pathParts[pathParts.length - 2];
						if (slug && slug !== "." && slug !== "content" && slug !== "routes" && slug !== "src") {
							basename = slug;
						}
					}
				}

				pathToAtUri.set(basename, postState.atUri);

				// Also add variations that might match HTML file paths
				// e.g., /blog/my-post, /posts/my-post, my-post/index
				const dirName = path.basename(path.dirname(filePath));
				// Skip route groups and common directory names
				if (dirName !== "." && dirName !== "content" && !(dirName.startsWith("(") && dirName.endsWith(")"))) {
					pathToAtUri.set(`${dirName}/${basename}`, postState.atUri);
				}
			}
		}

		if (pathToAtUri.size === 0) {
			log.warn(
				"No published posts found in state. Run 'sequoia publish' first.",
			);
			return;
		}

		log.info(`Found ${pathToAtUri.size} published posts in state`);

		// Scan for HTML files
		const htmlFiles = await glob("**/*.html", {
			cwd: resolvedOutputDir,
			absolute: false,
		});

		if (htmlFiles.length === 0) {
			log.warn(`No HTML files found in ${resolvedOutputDir}`);
			return;
		}

		log.info(`Found ${htmlFiles.length} HTML files`);

		let injectedCount = 0;
		let skippedCount = 0;
		let alreadyHasCount = 0;

		for (const file of htmlFiles) {
			const htmlPath = path.join(resolvedOutputDir, file);
			// Try to match this HTML file to a published post
			const relativePath = file;
			const htmlDir = path.dirname(relativePath);
			const htmlBasename = path.basename(relativePath, ".html");

			// Try different matching strategies
			let atUri: string | undefined;

			// Strategy 1: Direct basename match (e.g., my-post.html -> my-post)
			atUri = pathToAtUri.get(htmlBasename);

			// Strategy 2: Directory name for index.html (e.g., my-post/index.html -> my-post)
			if (!atUri && htmlBasename === "index" && htmlDir !== ".") {
				const slug = path.basename(htmlDir);
				atUri = pathToAtUri.get(slug);

				// Also try parent/slug pattern
				if (!atUri) {
					const parentDir = path.dirname(htmlDir);
					if (parentDir !== ".") {
						atUri = pathToAtUri.get(`${path.basename(parentDir)}/${slug}`);
					}
				}
			}

			// Strategy 3: Full path match (e.g., blog/my-post.html -> blog/my-post)
			if (!atUri && htmlDir !== ".") {
				atUri = pathToAtUri.get(`${htmlDir}/${htmlBasename}`);
			}

			if (!atUri) {
				skippedCount++;
				continue;
			}

			// Read the HTML file
			let content = await fs.readFile(htmlPath, "utf-8");

			// Check if link tag already exists
			const linkTag = `<link rel="site.standard.document" href="${atUri}">`;
			if (content.includes('rel="site.standard.document"')) {
				alreadyHasCount++;
				continue;
			}

			// Find </head> and inject before it
			const headCloseIndex = content.indexOf("</head>");
			if (headCloseIndex === -1) {
				log.warn(`  No </head> found in ${relativePath}, skipping`);
				skippedCount++;
				continue;
			}

			if (dryRun) {
				log.message(`  Would inject into: ${relativePath}`);
				log.message(`    ${linkTag}`);
				injectedCount++;
				continue;
			}

			// Inject the link tag
			const indent = "  "; // Standard indentation
			content =
				content.slice(0, headCloseIndex) +
				`${indent}${linkTag}\n${indent}` +
				content.slice(headCloseIndex);

			await fs.writeFile(htmlPath, content);
			log.success(`  Injected into: ${relativePath}`);
			injectedCount++;
		}

		// Summary
		log.message("\n---");
		if (dryRun) {
			log.info("Dry run complete. No changes made.");
		}
		log.info(`Injected: ${injectedCount}`);
		log.info(`Already has tag: ${alreadyHasCount}`);
		log.info(`Skipped (no match): ${skippedCount}`);

		if (skippedCount > 0 && !dryRun) {
			log.info(
				"\nTip: Skipped files had no matching published post. This is normal for non-post pages.",
			);
		}
	},
});

1	import * as fs from "fs/promises";
2	import { command, flag, option, optional, string } from "cmd-ts";
3	import { log } from "@clack/prompts";
4	import * as path from "path";
5	import { glob } from "glob";
6	import { loadConfig, loadState, findConfig } from "../lib/config";
7
8	export const injectCommand = command({
9	name: "inject",
10	description:
11	"Inject site.standard.document link tags into built HTML files",
12	args: {
13	outputDir: option({
14	long: "output",
15	short: "o",
16	description: "Output directory to scan for HTML files",
17	type: optional(string),
18	}),
19	dryRun: flag({
20	long: "dry-run",
21	short: "n",
22	description: "Preview what would be injected without making changes",
23	}),
24	},
25	handler: async ({ outputDir: outputDirArg, dryRun }) => {
26	// Load config
27	const configPath = await findConfig();
28	if (!configPath) {
29	log.error("No sequoia.json found. Run 'sequoia init' first.");
30	process.exit(1);
31	}
32
33	const config = await loadConfig(configPath);
34	const configDir = path.dirname(configPath);
35
36	// Determine output directory
37	const outputDir = outputDirArg \|\| config.outputDir \|\| "./dist";
38	const resolvedOutputDir = path.isAbsolute(outputDir)
39	? outputDir
40	: path.join(configDir, outputDir);
41
42	log.info(`Scanning for HTML files in: ${resolvedOutputDir}`);
43
44	// Load state to get atUri mappings
45	const state = await loadState(configDir);
46
47	// Generic filenames where the slug is the parent directory, not the filename
48	// Covers: SvelteKit (+page), Astro/Hugo (index), Next.js (page), etc.
49	const genericFilenames = new Set([
50	"+page",
51	"index",
52	"_index",
53	"page",
54	"readme",
55	]);
56
57	// Build a map of slug/path to atUri from state
58	const pathToAtUri = new Map<string, string>();
59	for (const [filePath, postState] of Object.entries(state.posts)) {
60	if (postState.atUri) {
61	// Extract slug from file path (e.g., ./content/blog/my-post.md -> my-post)
62	let basename = path.basename(filePath, path.extname(filePath));
63
64	// If the filename is a generic convention name, use the parent directory as slug
65	if (genericFilenames.has(basename.toLowerCase())) {
66	// Split path and filter out route groups like (blog-article)
67	const pathParts = filePath
68	.split(/[/\\]/)
69	.filter((p) => p && !(p.startsWith("(") && p.endsWith(")")));
70	// The slug should be the second-to-last part (last is the filename)
71	if (pathParts.length >= 2) {
72	const slug = pathParts[pathParts.length - 2];
73	if (slug && slug !== "." && slug !== "content" && slug !== "routes" && slug !== "src") {
74	basename = slug;
75	}
76	}
77	}
78
79	pathToAtUri.set(basename, postState.atUri);
80
81	// Also add variations that might match HTML file paths
82	// e.g., /blog/my-post, /posts/my-post, my-post/index
83	const dirName = path.basename(path.dirname(filePath));
84	// Skip route groups and common directory names
85	if (dirName !== "." && dirName !== "content" && !(dirName.startsWith("(") && dirName.endsWith(")"))) {
86	pathToAtUri.set(`${dirName}/${basename}`, postState.atUri);
87	}
88	}
89	}
90
91	if (pathToAtUri.size === 0) {
92	log.warn(
93	"No published posts found in state. Run 'sequoia publish' first.",
94	);
95	return;
96	}
97
98	log.info(`Found ${pathToAtUri.size} published posts in state`);
99
100	// Scan for HTML files
101	const htmlFiles = await glob("*/.html", {
102	cwd: resolvedOutputDir,
103	absolute: false,
104	});
105
106	if (htmlFiles.length === 0) {
107	log.warn(`No HTML files found in ${resolvedOutputDir}`);
108	return;
109	}
110
111	log.info(`Found ${htmlFiles.length} HTML files`);
112
113	let injectedCount = 0;
114	let skippedCount = 0;
115	let alreadyHasCount = 0;
116
117	for (const file of htmlFiles) {
118	const htmlPath = path.join(resolvedOutputDir, file);
119	// Try to match this HTML file to a published post
120	const relativePath = file;
121	const htmlDir = path.dirname(relativePath);
122	const htmlBasename = path.basename(relativePath, ".html");
123
124	// Try different matching strategies
125	let atUri: string \| undefined;
126
127	// Strategy 1: Direct basename match (e.g., my-post.html -> my-post)
128	atUri = pathToAtUri.get(htmlBasename);
129
130	// Strategy 2: Directory name for index.html (e.g., my-post/index.html -> my-post)
131	if (!atUri && htmlBasename === "index" && htmlDir !== ".") {
132	const slug = path.basename(htmlDir);
133	atUri = pathToAtUri.get(slug);
134
135	// Also try parent/slug pattern
136	if (!atUri) {
137	const parentDir = path.dirname(htmlDir);
138	if (parentDir !== ".") {
139	atUri = pathToAtUri.get(`${path.basename(parentDir)}/${slug}`);
140	}
141	}
142	}
143
144	// Strategy 3: Full path match (e.g., blog/my-post.html -> blog/my-post)
145	if (!atUri && htmlDir !== ".") {
146	atUri = pathToAtUri.get(`${htmlDir}/${htmlBasename}`);
147	}
148
149	if (!atUri) {
150	skippedCount++;
151	continue;
152	}
153
154	// Read the HTML file
155	let content = await fs.readFile(htmlPath, "utf-8");
156
157	// Check if link tag already exists
158	const linkTag = `<link rel="site.standard.document" href="${atUri}">`;
159	if (content.includes('rel="site.standard.document"')) {
160	alreadyHasCount++;
161	continue;
162	}
163
164	// Find </head> and inject before it
165	const headCloseIndex = content.indexOf("</head>");
166	if (headCloseIndex === -1) {
167	log.warn(` No </head> found in ${relativePath}, skipping`);
168	skippedCount++;
169	continue;
170	}
171
172	if (dryRun) {
173	log.message(` Would inject into: ${relativePath}`);
174	log.message(` ${linkTag}`);
175	injectedCount++;
176	continue;
177	}
178
179	// Inject the link tag
180	const indent = " "; // Standard indentation
181	content =
182	content.slice(0, headCloseIndex) +
183	`${indent}${linkTag}\n${indent}` +
184	content.slice(headCloseIndex);
185
186	await fs.writeFile(htmlPath, content);
187	log.success(` Injected into: ${relativePath}`);
188	injectedCount++;
189	}
190
191	// Summary
192	log.message("\n---");
193	if (dryRun) {
194	log.info("Dry run complete. No changes made.");
195	}
196	log.info(`Injected: ${injectedCount}`);
197	log.info(`Already has tag: ${alreadyHasCount}`);
198	log.info(`Skipped (no match): ${skippedCount}`);
199
200	if (skippedCount > 0 && !dryRun) {
201	log.info(
202	"\nTip: Skipped files had no matching published post. This is normal for non-post pages.",
203	);
204	}
205	},
206	});