Doesn't seem to support uploading a directory tree?

[mattj65817]

Is it me, or could this never possibly work to upload a tree of directories that is >1 level deep?

for (const dir of config.directories) {
        info(`--> Uploading directory ${dir} (excluded: ${excluded.includes(dir)})`);
        if (excluded.includes(dir)) continue;

        const globber = await createGlobPattern(dir, { followSymbolicLinks: config.followSymlinks });
        for await (const d of globber.globGenerator()) {
            const stats = await lstat(d);
            if (!stats.isDirectory()) {
                warning(`Path ${d} is not a directory, skipping!`);
                continue;
            }

            const files = await readdir(d).then((files) => files.map((s) => join(d, s)));
            for (const file of files) {
                debug(`Uploading file ${file} (from directory ${d})`);
                if (excluded.includes(file)) continue;

                await upload({
                    pathFormat: config.pathFormat,
                    partSize: config.partSize,
                    prefix: config.prefix,
                    bucket: config.bucket,
                    stream: createReadStream(file),
                    file: basename(file),
                    acl: config.objectAcl
                });

The issue appears to be in const files = await readdir(d).then((files) => files.map((s) => join(d, s)));.

1. We glob for directories matching the directories config.
2. We read each matched directory and assume every entry in it is a file (not a subdirectory) and try to upload it as a file.
3. If any entry is not a file, is a directory, we blow up with Error: failed to run noelware/s3-action: Error: EISDIR: illegal operation on a directory, read

I feel like I have to be missing something, I don't know how you get to a version 2.x without uncovering an error like this. It seems like a simple fix: after readdir(d), in files.map(), filter for files; in other words, filter out directories.

The documentation seems to state that this should work:

directories: List[String] (Required: No, Default: [])
The directories input indicates which directories should be uploaded to Amazon S3. The action will recursively get all the files in said directories

...but I don't see how that could possibly be the case.

I can submit a PR if needed, but it seems like I have to be missing something here. Surely somebody would've tried to upload a directory tree >= 2 levels deep by now?

[auguwu]

Probably not since we usually recurse a single directory but thanks for catching the issue! I wrote this issue without thinking about subdirectories but I'll provide a fix.

[mattj65817]

Probably not since we usually recurse a single directory but thanks for catching the issue! I wrote this issue without thinking about subdirectories but I'll provide a fix.

Thanks! I'm also seeing .html files uploaded with metadata calling them application/octet-stream, meaning browsers download them instead of displaying them, but I haven't checked the cause of that yet. I'll follow up on that when I can.

[auguwu]

In the latest revision (try master tag), I added { recursive: true } to fs.readdir's options, so this should do the trick.

Let me know if anything arise!

[mattj65817]

In the latest revision (try master tag), I added { recursive: true } to fs.readdir's options, so this should do the trick.

Let me know if anything arise!

Thank you! I worked around it by using find to dump out a list of individual files and passing that to s3-action, but I'll test this as soon as I get a chance.