Re: [Monotone-devel] Handling of empty/similar files

[graydon hoare]

Sebastien Pierre <address@hidden> writes:

> Hi all,

hello, and welcome!

> I was wondering how Monotone can deal with empty files, or files
> with the same content (same version code). For instance, in Python
> projects it is usual to place empty __init__.py files in each folder
> that represens a module:
> 
> Sources/Package
>  |-- Package
>  |   |-- __init.py__
>  |   |-- Module
>  |       |-- __init__.py
> 
> How does Monotone deal with this kind of files ?

in a design sense, they are treated the same as any other file. a
zero-byte file has a sha1 too. it is:

da39a3ee5e6b4b0d3255bfef95601890afd80709

in terms of implementation, a zero-byte file scans and imports ok, I
just checked. we may discover some logic errors in monotone related to
zero byte files, but those would be bugs.

in general, the storage system only keeps one copy of a particular
version -- more accurately one reconstruction path -- no matter how
many locations in a tree, or how many different tree versions, it
shows up in. the same is true of zero byte files.

there could be an inefficiency due to the storage system storing a
zero-length file at the *end* of a reconstruction chain of reverse
deltas, say if you store a bunch of zero-length files and then modify
one to have content and store it again. if we see this happening
often, and costing a lot, it could be worth some special treatment.

-graydon