Support for monorepos based on npm/Yarn Workspaces #1567

efokschaner · 2021-04-11T08:15:48Z

As mentioned in #1548 , and due to the fact that the previous solution to this stopped working with typedoc 0.20 I wanted to see if I could implement support for npm / Yarn Workspaces.

It still needs some automated tests as well as some design direction, but I just wanted to show how it looks, get the maintainers' initial thoughts, and see if this is a feature / mode that typedoc would be interested in having. (If not, I'd really appreciate guidance on how the same thing could be pulled off as a plugin?)

See the README for an overview of how this mode works. You can also see this version in action in the docs for this monorepo: https://efokschaner.github.io/laniakea/index.html

I'll leave some comments around the PR for some specific things I'd like to discuss.

Contrast with classic "entrypoint" style invocation

When passing --packages rather than entry points, we only document the single "main" entry point for each package found. This means we're only documenting the package's external api, and not all the modules within the package forming its implementation. This may be desirable, but some may still prefer to document each inner module within their packages. This also means currently we wouldn't support packages with multiple entry points (the kind which are designed for import/require using "packagename/submodule" syntax).

In addition, we only use the default "implicit" tsconfig discovery process for each package (ts.findConfigFile). This is an assumption that the package's tsc invocation isn't doing anything too fancy with its command line params or the location / name of the tsconfig.

There seems to be a theme of potentially needing configuration at the per-package level in order to deal with cases where people would need or want non-default behaviour. I could see a world perhaps where we eventually introduce support for a typedoc.json at the level of each individual package to deal with these kinds of things.

Inter-package references

One of the current limitations is that in this mode, typedoc sees inter-package references just like tsc does, as references to type declarations, rather than to the original source.

For example, see this reference to NominalType where it says it's defined in utils/dist/nominal-type.d.ts:5 rather than connecting it to the original source as it does here

vscode has had similar issues to this in the past with "Go to definition" taking you to .d.ts files (eg. microsoft/vscode#19672). Eventually this was solved via "declaration maps" and potentially the same thing could be used here to connect the inter-package references back to the original source by having typedoc use the declaration maps.

Tests

If there's interest in integrating this feature, I'm wondering what kind of tests would be preferred? My instinct is to begin with a few happy-path end-to-end regression tests, based on a couple of new sample projects in the examples folder to demonstrate the different ways one can use --packages, perhaps using the json converter output format to freeze the expected results?

Based on npm / Yarn workspaces, introducing the --packages option

efokschaner · 2021-04-11T08:18:05Z

src/lib/converter/converter.ts

@@ -19,6 +19,12 @@ import { IMinimatch } from "minimatch";
 import { hasAllFlags, hasAnyFlag } from "../utils/enum";
 import { resolveAliasedSymbol } from "./utils/symbols";

+export interface PackageProgram {


I'm not that happy with the name PackageProgram for this concept, any suggestions?

DocumentationEntrypoint perhaps? This is a hard thing to name...

efokschaner · 2021-04-11T08:21:28Z

src/lib/converter/converter.ts

@@ -134,11 +140,13 @@ export class Converter extends ChildableComponent<
     * Compile the given source files and create a project reflection for them.
     *
     * @param entryPoints the entry points of this program.
-     * @param program the program to document that has already been type checked.
+     * @param programs the programs to document, that have already been type checked.
+     * @param packages an array of packages (used in --packages mode)
     */
    convert(
        entryPoints: readonly string[],


I'm not that happy with the way in which I've plumbed the packages as a third param to convert where you're basically either using the first 2 params or the third. However I wasn't really sure if it would make sense to support passing non-package entry-points as well as packages simultaneously in to this system, or if these data structures could be unified earlier in the processing to make this less weird.

Looking for ideas here on better architecture.

I think the way I'd like to handle this is to make a breaking change, and accept an array of DocumentationEntrypoint, then the caller is responsible for turning an array of programs (in the entryPoint case) or packages into the appropriate structure.

Means this can't be released without a minor version bump, but I think this feature deserves one anyways - I'll take what I've done towards 0.21, merge it with this, and call the theme refactor 0.22 rather than making this wait.

efokschaner · 2021-04-11T08:22:57Z

src/lib/converter/converter.ts

+        entryPoints: readonly string[],
+        context: Context,
+        packages: PackageProgram[]
+    ) {
        const baseDir = getCommonDirectory(entryPoints);
        const entries: {


We do manage to unify the different sources of program / entry point in to the entries here.

efokschaner · 2021-04-11T08:23:56Z

src/lib/converter/converter.ts

@@ -269,7 +282,12 @@ export class Converter extends ChildableComponent<
            for (const program of context.programs) {
                const sourceFile = program.getSourceFile(file);
                if (sourceFile) {
-                    entries.push({ file, sourceFile, program });
+                    entries.push({
+                        entryName: getModuleName(resolve(file), baseDir),


getModuleName logic just moved up from below

efokschaner · 2021-04-11T08:25:37Z

src/lib/converter/converter.ts

        for (const entry of entries) {
            context.setActiveProgram(entry.program);
            entry.context = this.convertExports(
                context,
                entry.sourceFile,
-                entryPoints,
-                getModuleName(resolve(entry.file), baseDir)
+                entries.length === 1,


Moving entries.length === 1 up from below wasn't necessary, I had some potential changes in this code which I think I didn't need in the end, but left this as I found it simpler to understand than seeing the whole array passed.

efokschaner · 2021-04-11T08:28:02Z

src/lib/output/themes/DefaultTheme.ts

        const builder = new NavigationBuilder(
            project,
            project,
-            this.application.options.getValue("entryPoints").length > 1


"entryPoints" option no longer matches the number of modules we're documenting so I needed a change here. I think using project.getReflectionsByKind(ReflectionKind.Module).length is valid, but that might not be the number of modules we're going to actually document, so please let me know if this needs more investigation.

This does change the logic a bit, and unfortunately is considerably more expensive since it loops over every reflection in the project (both direct children and their children). A better check would be project.getChildrenByKind(ReflectionKind.Module).length !== 0

This suggestion seems to work 👍

efokschaner · 2021-04-11T08:28:52Z

src/lib/utils/package-manifest.ts

+ * @param obj the receiver of the hasOwnProperty method call
+ * @param prop the property to test for
+ */
+function hasOwnProperty<X extends {}, Y extends PropertyKey>(


This is a very generic utility function but I wasn't sure if it should live elsewhere.

Here is fine, if something else needs it it can always be moved.

efokschaner · 2021-04-11T08:30:34Z

examples/basic/.gitignore

@@ -0,0 +1 @@
+docs


Running typedoc on the examples folder locally, I needed to gitignore this folder.

Gerrit0 · 2021-04-19T03:34:14Z

Sorry I haven't had the time to look at this yet! It looks like a really neat idea. I should be able to look at it next weekend.

efokschaner · 2021-04-19T05:36:32Z

@Gerrit0 thanks for letting me know you saw this. Of course, take however long you wish.

AllanZhengYP · 2021-04-23T17:43:05Z

This is super useful, can we get this supported in next release?

Gerrit0

Looking good!

Gerrit0 · 2021-04-23T16:26:26Z

bin/typedoc

+    if (
+        app.options.getValue("entryPoints").length === 0 &&
+        app.options.getValue("packages").length === 0
+    ) {
        app.logger.error("No entry points provided");


I'd like to also mention packages here

Gerrit0 · 2021-04-24T15:07:38Z

src/lib/converter/converter.ts

@@ -19,6 +19,12 @@ import { IMinimatch } from "minimatch";
 import { hasAllFlags, hasAnyFlag } from "../utils/enum";
 import { resolveAliasedSymbol } from "./utils/symbols";

+export interface PackageProgram {


DocumentationEntrypoint perhaps? This is a hard thing to name...

Gerrit0 · 2021-04-24T15:09:00Z

src/lib/converter/converter.ts

@@ -134,11 +140,13 @@ export class Converter extends ChildableComponent<
     * Compile the given source files and create a project reflection for them.
     *
     * @param entryPoints the entry points of this program.
-     * @param program the program to document that has already been type checked.
+     * @param programs the programs to document, that have already been type checked.
+     * @param packages an array of packages (used in --packages mode)
     */
    convert(
        entryPoints: readonly string[],


I think the way I'd like to handle this is to make a breaking change, and accept an array of DocumentationEntrypoint, then the caller is responsible for turning an array of programs (in the entryPoint case) or packages into the appropriate structure.

Means this can't be released without a minor version bump, but I think this feature deserves one anyways - I'll take what I've done towards 0.21, merge it with this, and call the theme refactor 0.22 rather than making this wait.

Gerrit0 · 2021-04-24T15:13:26Z

src/lib/output/themes/DefaultTheme.ts

        const builder = new NavigationBuilder(
            project,
            project,
-            this.application.options.getValue("entryPoints").length > 1


This does change the logic a bit, and unfortunately is considerably more expensive since it loops over every reflection in the project (both direct children and their children). A better check would be project.getChildrenByKind(ReflectionKind.Module).length !== 0

Gerrit0 · 2021-04-24T15:14:26Z

src/lib/utils/package-manifest.ts

+ * @param obj the receiver of the hasOwnProperty method call
+ * @param prop the property to test for
+ */
+function hasOwnProperty<X extends {}, Y extends PropertyKey>(


Here is fine, if something else needs it it can always be moved.

Gerrit0 · 2021-04-24T15:21:18Z

src/lib/utils/package-manifest.ts

+        typeof packageJSON.workspaces === "object" &&
+        packageJSON.workspaces != null &&
+        hasOwnProperty(packageJSON.workspaces, "packages") &&
+        Array.isArray(packageJSON.workspaces.packages)


Gerrit0 · 2021-04-24T15:24:09Z

src/lib/utils/package-manifest.ts

+ * Finds the corresponding TS file from a transpiled JS file.
+ * The JS must be built with sourcemaps.
+ */
+function getTsSourceFromJsSource(


This doesn't handle inline source maps properly... probably fine to ignore that as a first pass.

So i was thinking of pulling in a proper sourcemap lib, rather than the quick homegrown approach I took, which would presumably handle all kinds of things better than i have.|

Unfortunately they tended to be asynchronous (rightly so), which would likely mean re-working some of typedoc's apis to be async, which is probably warranted but felt like a big change.

Yeah, TypeDoc's APIs are slowly becoming async... but it's happening on an as-needed basis, so if something doesn't require async yet, it doesn't use it. I think we just ignore this edge case for now.

Gerrit0 · 2021-04-24T15:26:22Z

src/lib/utils/package-manifest.ts

+): string | undefined {
+    const contents = readFile(jsPath);
+    const sourceMapPrefix = "//# sourceMappingURL=";
+    const searchResult = contents.search(


Am I missing some reason contents.indexOf("\n//# sourceMappingURL=") doesn't work?

Seems like it would work.

Gerrit0 · 2021-04-24T15:27:03Z

src/lib/utils/package-manifest.ts

+    }
+    // There's a pretty large assumption in here that we only have
+    // 1 source file per js file. This is a pretty standard typescript approach,
+    // but people might do interesting things with transpilation that could break this.


I don't think I care - at least for a first pass.

Gerrit0 · 2021-04-24T15:29:43Z

src/lib/application.ts

+            packageJsonPath,
+            packageJson
+        );
+        if (packageEntryPoint === undefined) {


Question - why is failing to find an entry point sometimes a fatal error and sometimes just a warning? Is one reason for failure more common/expected?

My razor here would be:
Ignore things which don't appear to be require-able node packages.
Fail on things which appear to be require-able node packages but are missing the necessary metadata for us to document.

I'll actually add the above as a comment in the code to guide maintainers.

The scenario which I saw so far, was just from running this across one of my own codebases where some of my packages were actually webpack browser bundles, not intended to be consumed as node.js modules, so there is no js entry point to be seen. This package for example: https://github.com/efokschaner/laniakea/tree/master/demos/balls/client

I fully anticipate we're going to see some surprising things in the wild w.r.t to packages and package structure for which this first pass at some default behaviour will not suffice. This comes back to my general remark above about potentially needing support for per-package typedoc config in the future, when we learn what people might need to tune.

Makes sense 👍
Per-package configuration might be difficult... excludeExternals for example changes the theme output... future work.

Gerrit0 · 2021-04-24T15:32:09Z

This is super useful, can we get this supported in next release?

I don't plan on holding this up once it's ready, will release as 0.21 and push the theme rework to 0.22.

Gerrit0 · 2021-04-24T15:34:34Z

For tests - the way I'd like to handle this is a couple small monorepo style projects under src/test/packages and either snapshotting the JSON like the converter tests do, or spot checking like the converter2 tests do. The examples folder should really have never been used for tests, I just haven't moved them out of that folder yet.

… in.

efokschaner · 2021-05-03T02:53:57Z

src/lib/utils/package-manifest.ts

+    let source = sourceMap.sources[0];
+    // If we have a sourceRoot, trim any leading slash from the source, and join them
+    // Similar to how it's done at https://github.com/mozilla/source-map/blob/58819f09018d56ef84dc41ba9c93f554e0645169/lib/util.js#L412
+    if (sourceRoot !== undefined) {


Source maps don't have to contain a sourceRoot and we handle that now. I also got this a little more conformant with the spec.

efokschaner · 2021-05-03T02:55:02Z

src/lib/utils/package-manifest.ts

+    // We can use require.resolve to let node do its magic.
+    // Pass an empty `paths` as node_modules locations do not need to be examined
+    try {
+        jsEntryPointPath = require.resolve(jsEntryPointPath, { paths: [] });


This is the other change in here. package.json's main values seem to be roughly as flexible as the require() paths themselves in terms of what actual file gets loaded.

efokschaner · 2021-05-03T02:55:55Z

package.json

@@ -67,7 +67,7 @@
  ],
  "scripts": {
    "pretest": "node scripts/copy_test_files.js",
-    "test": "nyc --reporter=html --reporter=text-summary mocha --timeout=10000 'dist/test/**/*.test.js'",
+    "test": "nyc --reporter=html --reporter=text-summary mocha --timeout=10000 'dist/test/**/*.test.js' --exclude 'dist/test/packages/**'",


Because the /packages/ folder contains a non-trivial amount of sample code we need to avoid accidentally picking up tests in there

efokschaner · 2021-05-03T03:12:59Z

src/lib/utils/options/readers/typedoc.ts

            try {
+                // The "packages" option is an array of paths and should be interpreted as relative to the typedoc.json


This came up in testing because the tests cwd is not inside the project we're operating on, and without this the paths in the typedoc.json are interpreted relative to cwd, rather than to the file in which they appear.

This actually seems like something you might want to change for all options which are filepaths but could appear in the typedoc.json.

Yes! I completely agree, #1442 tracks this, but last time I went after it the approach I took failed.

efokschaner · 2021-05-03T03:16:20Z

scripts/copy_test_files.js

+            );
+            if (npmMajorVer < 7) {
+                throw new Error(
+                    "npm version must be at least 7, version installed is " +


Installing the monorepo's deps (and gluing the inter-package deps together) requires either lerna, or yarn, or npm>7. I opted for the npm approach as I figured it's the most "vanilla" despite having been around for the least amount of time. This could run in to issues under some of your CI configs however so let's see if anything breaks and decide what to do.

efokschaner · 2021-05-03T03:17:20Z

src/test/packages/typedoc-single-package-example/package.json

+    "example"
+  ],
+  "author": "",
+  "license": "Apache-2.0",


I copied the same license as typedoc itself, not that it should matter much for this little island of code.

efokschaner · 2021-05-03T03:18:53Z

src/test/packages/typedoc-single-package-example/package.json

+    "test": "node ."
+  },
+  "devDependencies": {
+    "typescript": "^4.2.4"


Technically this package should depend on typedoc? I never had to install it because the tests bring their own typedoc.
It gets a bit weird though as if treated in isolation this would pull in typedoc from the internet. What you actually have here is a monorepo XD where you want packages developed in the same source tree to use each other's development copies.

efokschaner · 2021-05-03T03:25:33Z

I recommend reviewing the newest changes commit by commit.

I was fairly happy with the outcome of your DocumentationEntrypoint suggestions, in this commit 23db9a3

In 998e610 I took a slightly elaborate approach to the testing, but one that gave me pretty good results and already lead to me improving the new behaviour in a couple of ways. Let me know if this test approach seems silly.

I went looking for good examples of a skeleton of monorepo setup and I found this repo to be a well built example: https://github.com/NiGhTTraX/ts-monorepo . Rather than vendor that in to the tests, or embroil myself in git submodules, I decided to clone the repo during scripts/copy_test_files.js. It adds a minute or so to clone and install and build things. You can still iterate on the tests quickly just running nyc + mocha directly.

I actually forked the repo and added typedoc support (efokschaner/ts-monorepo@49a565e), and so the tests pull from my fork (with a pinned commit sha). You can see the docs typedoc currently generates here (https://efokschaner.github.io/ts-monorepo/)

I was actually considering approaching the author of ts-monorepo to see if they would like to include typedoc setup as part of their example, which would then eliminate my own fork. If you're happy with ts-monorepo as a test case but don't like the connection to my fork we can do a couple of other things such as, fork it under your own org / vendor the repo / clone from the upstream and patch the typedoc support in after clone.

Let me know how things look, what else to change.

As we get a little closer to merging this feature, I was wondering if it would make sense to pitch the monorepo support as a little more experimental than the rest of the typedoc featureset, in the readme/docs? Just to manage expectations as it's likely to discover a fair bit of complexity.

PS. Github workflows just introduced "approval for first-timers", which is now pending on this PR.

efokschaner · 2021-05-03T03:28:12Z

src/test/packages.test.ts

+        ok(project, "Failed to convert");
+        const result = app.serializer.projectToObject(project);
+        ok(result.children !== undefined);
+        strictEqual(


The validation I'm doing in this and other cases is pretty minimal atm, I mainly wanted to focus on the new behaviour around discovering package entrypoints as I assume most other behaviour is already tested elsewhere. LMK if there are other things you believe we should check here though.

Gerrit0 · 2021-05-15T16:39:07Z

I need to pull this down and play with it some more, but overall I'm really happy with this. I think I'll probably pull the monorepo tests into this repo rather than downloading them, since keeping tests fast is really important to me (TypeDoc's tests used to take 3 minutes to run, which pretty much meant they were useless for quickly verifying that everything worked as expected... getting it down much lower was a huge productivity boost)

Gerrit0 · 2021-06-06T17:37:23Z

If you're looking to try this out - it has been published under 0.21.0-beta.2

Monkatraz · 2021-06-06T21:22:43Z

I gave this a shot, and I'm having an issue with my particular monorepo. It is a PNPM setup, but manually pointing packages to a glob of my package folder seems fine. The issue isn't really with that, but instead with how my packages work. My packages are uncompiled, and their main property points to index.ts, not a compiled index.js file. This doesn't seem to work, as I get a xyz does not contain a sourceMappingURL.

Is this an unsupported edge case?

efokschaner · 2021-06-08T08:21:20Z

Hey @Monkatraz, it's not a use case I'd considered, but it sounds pretty easy to support.

I can push another commit to this PR or we can do a separate PR based on @Gerrit0 's preference.

efokschaner · 2021-06-08T08:50:03Z

@Gerrit0 regarding your comment about vendoring the monorepo code, that probably won't significantly improve the time to bootstrap the tests as the clone is outweighed by the time to install deps and build that repo, which is what's required to have the js generated for us to have sourcemaps to find the TS entry points.

I agree with your objective to keep the test cycle fast of course.

One solution to that could be to not wipe+install+rebuild the codebase used in the tests each time the full test script is called. This has some downsides in that the first test run still takes the full time, and the test run is no longer from a completely clean state by default. The solution I used to this locally was to just separate my use of test setup and test iteration. I ran the npm test command once to get the copy_test_files pretest setup done once and then just ran the nyc mocha command directly to iterate without all the re-setup. So I intentionally opted in to the not-from-scratch test iteration workflow.

So I could see 3 possible approaches here:

Vendor the monorepo in its already built state so copy_test_files has no install + build work to do. I'm not sure if this would require committing all its node_modules, I think tsc needs them for typedoc to work?
Change copy_test_files to not start with a clean slate if the dir appears to be setup. Slow on first run but faster after.
Better document + support the "full test with setup" vs "iterate on tests without re-setup" approach. Slow on first run but faster after and explicit about opting in to non-clean-slate test runs.

Do you have other options in mind? What do you think makes most sense?

Unnvaldr · 2021-06-09T01:00:23Z

I've noticed that assumed base directory for packages is a location of the TypeDoc config (which is different from entryPoints approach), shouldn't it be uniform with current behaviour until it can be changed simultaneously?

I agree with Monkatraz, we should get support for main and also types, when they're pointing to TypeScript files.

Also, is there any reason for defining getEntrypointsForPackages as function, instead of making it a method in Application class?

edit: Found solutions to both matters

efokschaner · 2021-06-09T03:47:29Z

Hey @lhoerion , regarding the paths relative to typedoc.json:

shouldn't it be uniform with current behaviour until it can be changed simultaneously?

I can see the reasoning but given that this is an entirely new config option, unlikely to be used simultaneously with the entryPoints option, and that the way this one works is more likely to match people's default assumptions about how a path in a config file works, I'd advocate it do it the better way from the outset.
I defer to @Gerrit0 's preference on this of course. We'll need some kind of adjustment for the functional tests if we change this this as the tests operate with CWD !== typedoc.json folder.

Can you elaborate on how you see the package json types field interacting with the entrypoint-discovery logic? I hadn't considered that avenue.

getEntrypointsForPackages is not a member function just because it's a taste I have which minimises the surface area between data and code. Scott Meyers explains it nicely here: https://www.drdobbs.com/cpp/how-non-member-functions-improve-encapsu/184401197

Unnvaldr · 2021-06-09T13:00:22Z

Can you elaborate on how you see the package json types field interacting with the entrypoint-discovery logic? I hadn't considered that avenue.

If someone wish to document type definitions package, then usually they won't have main in package.json, only types.

I can see the reasoning but given that this is an entirely new config option, unlikely to be used simultaneously with the entryPoints option, and that the way this one works is more likely to match people's default assumptions about how a path in a config file works, I'd advocate it do it the better way from the outset.
I defer to @Gerrit0 's preference on this of course. We'll need some kind of adjustment for the functional tests if we change this this as the tests operate with CWD !== typedoc.json folder.

True, I'm also in favour of leaving it as-is, but we've to document this difference then.

Gerrit0 · 2021-06-13T15:16:04Z

Do you have other options in mind? What do you think makes most sense?

When I pulled this into the beta branch, what I ended up doing was pulling the project into this repo and also committing built files. The additional dependencies didn't seem to add any value, so the tests are very small now.

I defer to @Gerrit0 's preference on this of course.

I like the behavior it has - and have wanted to change the rest of the options that have paths to use this same behavior, just haven't gotten around to it yet... going to try it this morning.

efokschaner · 2021-06-13T19:44:59Z

@Gerrit0 thanks for your support integrating this code and getting the tests in to a good state.

I just wanted to check if you have any expectations from me about anything remaining to fix or change for this PR. If not, feel free to close this PR.

If we were to implement @Monkatraz 's use case here #1567 (comment) should I make that PR to the beta branch?

Unnvaldr · 2021-06-13T20:26:17Z

@Gerrit0 thanks for your support integrating this code and getting the tests in to a good state.

I just wanted to check if you have any expectations from me about anything remaining to fix or change for this PR. If not, feel free to close this PR.

If we were to implement @Monkatraz 's use case here #1567 (comment) should I make that PR to the beta branch?

I think that my PR covered that case as well

efokschaner · 2021-06-13T23:21:42Z

@lhoerion ah yes! Ty. Does it work to test for the file being .ts after it's passed to require.resolve?

Unnvaldr · 2021-06-14T11:21:43Z

@lhoerion ah yes! Ty. Does it work to test for the file being .ts after it's passed to require.resolve?

Yes it works, that way .ts/.js extension doesn't need to be explicitly provided in main property.
Some packages in multi-package test were changed to also account for that.

Gerrit0 · 2021-06-15T03:16:51Z

I just wanted to check if you have any expectations from me about anything remaining to fix or change for this PR. If not, feel free to close this PR.

Nothing else from me! I've been leaving it open to remind myself to give you lots of thanks in the release notes for 0.21.

AllanZhengYP · 2021-06-17T07:49:09Z

Hi @efokschaner I gave the feature a try, I find it doesn't support the inline source option in getTsSourceFromJsSource(). I got no such file exception because it tries to open the encoded source code like:

data:application/json;base64,eyJ2ZXJzaW9uIjozLCJmaWxlIjoiaW5kZXguanMiLCJzb3VyY2VSb290IjoiIiwic291cmNlcyI6WyIuLi8uLi9zcmMvaW5kZXgudHMiXSwibmFtZXMiOltdLCJtYXBwaW5ncyI6Ijs7O0FBQUEsNERBQWtDO0FBQ2xDLHdEQUE4QiIsInNvdXJjZXNDb250ZW50IjpbImV4cG9ydCAqIGZyb20gXCIuL0Fib3J0Q29udHJvbGxlclwiO1xuZXhwb3J0ICogZnJvbSBcIi4vQWJvcnRTaWduYWxcIjtcbiJdfQ==

Another question is how to show the README page of sub packages.

Thank again for your work!

Unnvaldr · 2021-06-17T10:49:59Z

Hi @efokschaner I gave the feature a try, I find it doesn't support the inline source option in getTsSourceFromJsSource(). I got no such file exception because it tries to open the encoded source code like:
data:application/json;base64,eyJ2ZXJzaW9uIjozLCJmaWxlIjoiaW5kZXguanMiLCJzb3VyY2VSb290IjoiIiwic291cmNlcyI6WyIuLi8uLi9zcmMvaW5kZXgudHMiXSwibmFtZXMiOltdLCJtYXBwaW5ncyI6Ijs7O0FBQUEsNERBQWtDO0FBQ2xDLHdEQUE4QiIsInNvdXJjZXNDb250ZW50IjpbImV4cG9ydCAqIGZyb20gXCIuL0Fib3J0Q29udHJvbGxlclwiO1xuZXhwb3J0ICogZnJvbSBcIi4vQWJvcnRTaWduYWxcIjtcbiJdfQ==

Hi @AllanZhengYP, in beta branch that case is already handled, so you should try out latest beta release 0.21.0-beta.4

typedoc/src/lib/utils/package-manifest.ts

Line 137 in 90cd4a0

if (sourceMapURL.startsWith("data:application/json;base64,")) {

Gerrit0 · 2021-06-18T02:44:03Z

This has been fully released in 0.21.0, thanks again!

efokschaner added 2 commits April 10, 2021 23:51

feat: Support for monorepos

cebd113

Based on npm / Yarn workspaces, introducing the --packages option

Merge branch 'master' of github.com:TypeStrong/typedoc

21cacd2

efokschaner commented Apr 11, 2021

View reviewed changes

This was referenced Apr 23, 2021

Support for ignoreCompilerErrors #1403

Closed

Exports in ambient module don't get included in certain case #1571

Closed

Gerrit0 requested changes Apr 24, 2021

View reviewed changes

Gerrit0 mentioned this pull request Apr 24, 2021

Version 0.22 Tracking #1543

Closed

efokschaner added 5 commits April 30, 2021 17:31

Merge branch 'master' of github.com:TypeStrong/typedoc

ef1d9f7

Incorporate PR feedback. Reworks Converter convert() interface.

23db9a3

Entrypoint detection fixes for compatibility with ts-monorepo

fcc74b1

Interpret the packages option relative to the typedoc.json it appears…

50bc055

… in.

Add some tests for the --packages option

998e610

efokschaner commented May 3, 2021

View reviewed changes

efokschaner changed the title ~~[WIP Feedback Requested] Support for monorepos based on npm/Yarn Workspaces~~ Support for monorepos based on npm/Yarn Workspaces May 3, 2021

efokschaner mentioned this pull request May 21, 2021

Create an example on how to use typedoc within a Monorepo (lerna) #1548

Closed

Gerrit0 mentioned this pull request Jun 5, 2021

Add support for TypeScript 4.3 #1589

Closed

Nokel81 mentioned this pull request Jun 7, 2021

Upgrade typedoc lensapp/lens#2940

Merged

Unnvaldr mentioned this pull request Jun 9, 2021

Add support for TS entry points in monorepo configuration #1596

Merged

Gerrit0 closed this Jun 18, 2021

efokschaner mentioned this pull request Jul 28, 2021

When using packages allow allow linking to types between different packages #1612

Closed

		try {
		// The "packages" option is an array of paths and should be interpreted as relative to the typedoc.json

Support for monorepos based on npm/Yarn Workspaces #1567

Support for monorepos based on npm/Yarn Workspaces #1567

Conversation

efokschaner commented Apr 11, 2021 • edited Loading

Contrast with classic "entrypoint" style invocation

Inter-package references

Tests

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

efokschaner Apr 11, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

efokschaner May 2, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Gerrit0 commented Apr 19, 2021

efokschaner commented Apr 19, 2021

AllanZhengYP commented Apr 23, 2021

Gerrit0 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

efokschaner May 2, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Gerrit0 commented Apr 24, 2021

Gerrit0 commented Apr 24, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

efokschaner May 3, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

efokschaner May 3, 2021 • edited Loading

Choose a reason for hiding this comment

efokschaner May 3, 2021 • edited Loading

Choose a reason for hiding this comment

efokschaner commented May 3, 2021 • edited Loading

efokschaner May 3, 2021 • edited Loading

Choose a reason for hiding this comment

Gerrit0 commented May 15, 2021

Gerrit0 commented Jun 6, 2021

Monkatraz commented Jun 6, 2021

efokschaner commented Jun 8, 2021

efokschaner commented Jun 8, 2021

Unnvaldr commented Jun 9, 2021 • edited Loading

efokschaner commented Jun 9, 2021 • edited Loading

Unnvaldr commented Jun 9, 2021 • edited Loading

Gerrit0 commented Jun 13, 2021

efokschaner commented Jun 13, 2021

Unnvaldr commented Jun 13, 2021

efokschaner commented Jun 13, 2021 • edited Loading

Unnvaldr commented Jun 14, 2021

Gerrit0 commented Jun 15, 2021

AllanZhengYP commented Jun 17, 2021

Unnvaldr commented Jun 17, 2021

Gerrit0 commented Jun 18, 2021

efokschaner commented Apr 11, 2021 •

edited

Loading

efokschaner Apr 11, 2021 •

edited

Loading

efokschaner May 2, 2021 •

edited

Loading

efokschaner May 2, 2021 •

edited

Loading

efokschaner May 3, 2021 •

edited

Loading

efokschaner May 3, 2021 •

edited

Loading

efokschaner May 3, 2021 •

edited

Loading

efokschaner commented May 3, 2021 •

edited

Loading

efokschaner May 3, 2021 •

edited

Loading

Unnvaldr commented Jun 9, 2021 •

edited

Loading

efokschaner commented Jun 9, 2021 •

edited

Loading

Unnvaldr commented Jun 9, 2021 •

edited

Loading

efokschaner commented Jun 13, 2021 •

edited

Loading