Re: [PATCH 02/11] Factor out and export large blob writing code to arbitrary file handle

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 27, 2012, 17:29 UTC
Message-ID: <7v4nucb2xl.fsf@alter.siamese.dyndns.org>
In-Reply-To: <1330329315-11407-3-git-send-email-pclouds@gmail.com>

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

Show 5 quoted lines

> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  cache.h |    3 +++
>  entry.c |   39 ++++++++++++++++++++++++++-------------
>  2 files changed, 29 insertions(+), 13 deletions(-)

It was the goal of the original streaming output topic to helping more callers stream the data out directly from the object store in order to reduce memory pressure, and this series is very much in line with its spirit.

The static version of streaming_write_entry() in entry.c was very specific to writing out an index entry out to the working tree, and it made perfect sense to have the function in that file, but its interface was limited to the original context the function was used in.

The whole point of your refactoring in this patch is to make it available for callers outside that original context; e.g. archive that finds blob SHA-1 from a tree and writes the blob out to its standard output. They should not have to work with an API that takes a cache-entry and writes to a working tree file. And your result is much more generic.

So I think the external declaration and the definition should move to a more generic place, namely streaming.[ch]. It does not belong to entry.c anymore.

Thanks for working on this.