🐇 There are two errors. One resulted from forgetting sudo when installing dependencies. The other was the incorrect name for the OpenSSL library. It should be libssl-dev instead of openssl-dev. Maybe we can link the command list in the docs.

🐢 Oops, yes, we should at least keep that in mind.

🐇 I’m checking the most popular crates. getrandom retrieves a random number from the system. It looks much lighter than the rand crate.

🦊 We can use once_cell to initialize XvcEntityCounter. We currently use Once for this purpose.

🐇 I don’t think it will provide any better features.

🦊 For that case, yes, no better features. But the interface is something like:

impl<T> OnceCell<T> {
    const fn new() -> OnceCell<T> { ... }
    fn set(&self, value: T) -> Result<(), T> { ... }
    fn get(&self) -> Option<&T> { ... }
}

And this makes, for example, working with XvcRoot much easier. We are passing Arc<RwLock<XvcRootInner>>> everywhere. This is a heavy price when we only use it in a read-only manner. We can prevent most of these, when we use a read lock, by using OnceCell.

🐇 XvcConfig can benefit from this as well. We don’t update the config during runs.

🐢 Why do we want to assign it, though? We currently have a config field in XvcRootInner, and we get a reference to it with the config() method.

🐇 Ok. Let’s skip this for now. No need to worry before measuring the performance impact.

To make debugging easier, we can add file deletions in the ./start-readme script to the notebook itself.

Instead, I’ll run my watcher run-after-commit.sh to ensure that the files are deleted after commits.

That may also work; you can use both as well.

I noticed cli-opts pass --no-system-config etc. by default. Let’s deal with this first.

I fixed that.

I’m testing whether we are in the directory that we should be in with xvc.root("--absolute"), but it looks like it’s not possible to pass a string argument to the command.

Let me take a look at this.

It looks like there have been some changes in optional parameter handling in PyO3. You may need to deal with keyword arguments with decorators.

Now, let’s try xvc.file().track() once more with dir-0001/.

It seems to be working now. With an existing directory, it doesn’t show an error.

What does xvc.file().list() show?

It shows a single string with \n in it. It looks like we need to handle this in the output thread.

I added a replace("\\n", "\n") to output_str at the end, but it didn’t make a difference. I tested in the notebook with list_result.split("\n"), and these \n characters are indeed CRLF. So Jupyter shows CRLF in strings with \n, and this is not something we should try to fix, I think.

You can search how to show \n characters in a Jupyter notebook with CRLF.

The output string should be fed into repr, as in:

# Define a string with CRLF characters
text_crlf = "Hello\r\nWorld\r\nThis is a test string."

# Use repr to show the \r\n characters explicitly
print(repr(text_crlf))

I tested this, and GPT misleads. repr is when you want to show \n, not vice versa. When I print(list_result), it prints the results properly.

It looks like we don’t need to make this a priority now. We can tell the user in the notebook that the commands are intentionally returning strings, and they can process or print them however they want.