Exclude/filter documents with specific frontmatter (i.e. draft: true)

[spryzone]

It would be great to exclude or filter based on frontmatter, not only glob pattern. This means we could handle drafts at the content collection level, and avoid writing JavaScript filters for each route.

e.g.:

const posts = defineCollection({
    name: "posts",
    directory: "src/content/posts",
    include: "**/*.md",
    filter: (frontmatter) => frontmatter.published === true,
    schema: (z) => ({
        title: z.string(),
        description: z.string(),
        image: z.string(),
    }),
    transform: async (document, context) => {
        const html = await compileMarkdown(context, document);
        return {
            ...document,
            html,
        };
    },
});

[sdorra]

You don't need to write code for each route. Simply create a file and define exports for all your filters, such as posts.ts:

import { allPosts } from "content-collections";

export const drafts = allPosts.filter(p => !p.published);
export const published = allPosts.filter(p => p.published);

Then, import the filtered collections wherever you need them:

import { published } from "./posts";

It's similar to sorting. Please take a look at the docs. We might also consider adding filtering to the documentation.

[spryzone]

Thanks I'll do something like this. I just thought it would be better to skip processing these files, instead of processing them and then filtering them out.

[sdorra]

I understand, but I can imagine many use cases where someone would want to filter based on a computed property (the result of the transform function). If we introduce a filter that only applies to schema properties, we would need to document two filtering methods, which could be confusing. Therefore, I prefer to keep the API small and easy to understand.

If you're concerned about performance, the markdown conversion is cached, so it is computed only once and should be quite fast.

[Minhir]

@sdorra I would add another concern here: bundling. It might be that you don't want to bundle certain docs not only because of performance, but also due to security issues.

For example, you can have multiple access levels in your docs. In such a case, I want to be able to define multiple collections from one matching pattern:

const publicPosts = defineCollection({
    name: "posts",
    directory: "src/content/posts",
    include: "**/*.md",
    filter: (frontmatter) => frontmatter.level === "public"
});

const privatePosts = defineCollection({
    name: "posts",
    directory: "src/content/posts",
    include: "**/*.md",
    filter: (frontmatter) => frontmatter.level === "private"
});

Or it could be some post-filtering functionality like:

const posts = defineCollection({
    name: "posts",
    directory: "src/content/posts",
    include: "**/*.md",
});

const publicPosts = createSubCollection(posts, (frontmatter) => frontmatter.level === "public");
const privatePosts = createSubCollection(posts, (frontmatter) => frontmatter.level === "private");

The key point is to handle this at the bundling step, not at runtime.

[sdorra]

This is a very good argument. I need to consider the API, but I will add the feature. @Minhir, can you open an issue? I'm currently on vacation without my laptop, and I want to make sure I don't forget about it.

[Minhir]

Issue #620