How to use uv_fs_read to read a single line from an input file?

[HallofFamer]

I am trying to use libuv to add asynchronous file IO for an experimental programming language, and I am a bit confused about how to use uv_fs_read. At this point I have figured out how to read to the end of the file, or read a certain number of characters into buffer. However, it seems that to read a single character(like fget) or a single line(like fgets) is not possible. The former is not very useful anyway, as it doesnt make much sense to read a single char asynchronously. But the latter is a functionality that Id like to implement in the PL, and I have no idea how or if it is even possible to do this.

So here is my question, is it possible to use uv_fs_read to read a single line from an input file, if the position of the linebreak characters are not unknown ahead of time? If not, is there a workaround to this? Thanks.

[saghul]

You would need to read your til you find the line break.

Since uv_fs_read works with offsets, you can then resume reading where you left off.

[bnoordhuis]

No, or at least not directly, because it's not a FILE but a file descriptor (or a HANDLE on Windows.)

You can pass it to fdopen() however (_open_osfhandle + _fdopen on Windows, I think.)

[HallofFamer]

@bnoordhuis I see, thx for your answer. I have another question though, as I am aware of the fact that uv_fs_read accepts an offset as argument, and when -1 is supplied it uses the current offset. So is there a way to get the current offset value, kinda similar to how ftell works? Or is this not possible and we just have to keep track of the offset using a variable if needed?

[bnoordhuis]

Rule of thumb: always specify the position unless you're reading from a file where positional I/O doesn't work (like /dev files, pipes, append-only files, etc.)

Why: there's no telling what the actual file position is with concurrent readers/writers. That's also why there's no ftell/lseek equivalent in libuv.

[HallofFamer]

@bnoordhuis thx for your help and I am able to get libuv fs operations to work mostly on my code. On the other hand, I wonder what would be the recommended way to read a file to the end using libuv with callback? I heard you mentioned that using `uv_fs_stat’ may be able to give some idea of how much size is needed for the buffer, though there are caveats with this approach. What if the buffer size turns out not enough for reading till the end?

[bnoordhuis]

You read again (and again, and again) until you reach EOF. :-)

It's in general a good idea to take a stream-like approach and put some upper limit on the sizes of the chunks you read. You don't want to be slurping a 1 TB file all into memory.