docs: `fs.write()` variables are not clear

[jorangreef]

The docs for fs.write() state:

offset and length determine the part of the buffer to be written.

offset is clear, but one must assume that length has the same meaning as per buffer.slice(offset, length), i.e. that length is not relative to offset, but rather an absolute index into buffer as is expected (e.g. string.slice(offset, length)). Therefore, if one wanted to write 5 bytes from offset 12 at position 1 in the file, one would use offset=12, length=12+5.

... (err, written, buffer) where written specifies how many bytes were written from buffer.

This is not clear.

I would probably guess that written is a quantity number, i.e. if the first call to fs.write() returned written=2, then I would think that a count of 2 bytes had been written. I would therefore expect that fs.write(fd, buffer, offset=12, length=17, position=1) would return written=5 if everything was written out, and less than 5 if a partial write occurred.

What actually happens is that written=17 when the write is fully written out. This is very surprising. Furthermore, if written now represents an absolute index, is it an absolute index into the source buffer or an absolute index into the target file? If it is an absolute index into the target file, is it relative to position or not? It seems that written is actually then an absolute index relative to the source buffer?

Why is buffer returned in the write callback? Is this the same buffer I passed, or a slice? The docs should also make that clear.

[jorangreef]

I think it's actually the other way round, length is not an absolute index into source buffer but a quantity relative to offset. Is that right?

[jorangreef]

The docs for fs.read() are more clear:

length is an integer specifying the number of bytes to read.

Perhaps it would be better if (for both methods) length was renamed to size to differentiate it from the typical meaning of length and to imply quantity as opposed to absolute index? And if written was renamed to bytesWritten to bring it in line with fs.read()?

[jalafel]

Hi! Looking to pick up this issue. Seems relatively easy enough! From my understanding, there is just an ambiguity with length, wherein traditionally, one might assume that it's the length of a buffer beginning from the offset index. Whereas, it actually denotes the length of the number of bytes to read.

It'd make sense to change the word choice to size, then, and expand briefly on the variable for the fs.write() documentation.

(Also changing written to bytesWritten for consistency with the documentation on fs.read().)

Let me know if there is any misunderstanding on my part!

[jorangreef]

Thanks for taking it up @jessicaquynh!

[sam-github]

I'm confused here, you start off stating that you think length is used to describe offsets... which would indeed be weird (as would size!), but then comment later that you realize it was not being used that way...

It seems to me that some of the documentation is indeed not clear, but it doesn't seem to me that changing length to size makes any difference. You may think they imply some kind of subtle difference, but I don't have that impression. And if you believe the difference between size and length is so clear, perhaps you can describe the difference clearly?

In the absence of a clear and convincing description of why size and length mean different things, I think doc improvements should be focussed on the text which describes the argument's behaviour, not renaming the arguments.

[jalafel]

@sam-github Continuing from your message from the PR.

About the commit message: That's my bad. I was thinking back to Java, where length usually denotes the data structure (array, list, string) size + 1, counting from 0, thus the index reference. Whereas size would count the items from 1, and denotes the number of items/bits. I agree that there doesn't seem to be a difference in Javascript, and size is not oft used, or rather, used synonymously with length.

I'd be happy to just clarify in the documentation, as opposed to switching out the word choice.

[sam-github]

I didn't know that about java. Yes, I think there is much you can improve in making the text clear, but all the changes in fs don't help, and worse, will make future changes to fs un-backportable to LTS branches. I don't think the length->size change is worth it, but I like your other changes. Sorry, it was probably a lot of work going through fs like that! :-(

[jalafel]

That is, if I recall correctly from my classes at uni. It may have been specific only to DLL and SLL.

And it's not a problem at all, it wasn't any work :) I will attempt to make the documentation clearer, rather than changing the syntax!