From: Abhradeep Chakraborty <chakrabortyabhradeep79@gmail.com>
To: Junio C Hamano <gitster@pobox.com>
Cc: Abhradeep Chakraborty via GitGitGadget <gitgitgadget@gmail.com>,
git@vger.kernel.org, Taylor Blau <me@ttaylorr.com>,
Kaartic Sivaram <kaartic.sivaraam@gmail.com>,
Derrick Stolee <derrickstolee@github.com>
Subject: Re: [PATCH 3/5] roaring: teach Git to write roaring bitmaps
Date: Fri, 30 Sep 2022 21:53:45 +0530 [thread overview]
Message-ID: <CAPOJW5yxRETdVk014gQYFud9_Nrt+OQGSVNQ8Pw2wDEMMFMm1Q@mail.gmail.com> (raw)
In-Reply-To: <xmqqczbdl6wl.fsf@gitster.g>
On Fri, Sep 30, 2022 at 11:51 AM Junio C Hamano <gitster@pobox.com> wrote:
>
> "Abhradeep Chakraborty via GitGitGadget" <gitgitgadget@gmail.com>
> writes:
>
> > From: Abhradeep Chakraborty <chakrabortyabhradeep79@gmail.com>
> >
> > Roaring bitmaps are said to be more efficient (most of the time) than
> > ewah bitmaps. So Git might gain some optimization if it support roaring
> > bitmaps. As Roaring library has all the changes it needed to implement
> > roaring bitmaps in Git, Git can learn to write roaring bitmaps. However,
> > all the changes are backward-compatible.
> >
> > Teach Git to write roaring bitmaps.
>
> That is way underexplained. At least cover what the plans are, so
> that readers do not have to ask these questions:
>
> * When is the choice of bitmap type is made? Is it fixed at
> repository initialization time and once chosen other kinds cannot
> be used?
>
> * Is the bitmap file self describing? How does a reader know
> between ewah and roaring codepaths to use to read a given bitmap
> file? Is there enough room for extending the set of bitmap
> formats, or we cannot add other formats easily?
Hey Junio,
First of all, sorry that the next version is taking so much time to
land. We have a festival ("Durga Puja"; it is the biggest festival for
Bengalis) going on here now. So I am not that active.
I will explain briefly in the next version.
>
> Do you really need the global variable that holds the bitmap type?
>
> Wouldn't it be easier to write code that needs to deal with both
> types (e.g. in a repository with existing ewah bitmap, you want to
> do a repack and index the result using the roaring bitmap) if you
> passed the type through the callchain as a parameter?
I didn't want to go for "passing the type through the callchain as a
parameter" because that would cause changes to every affected function
definition. I found the "global variable" approach simpler for this
reason. Here we have to initialize the type once and the affected
functions will work accordingly.
If you like the "callchain" approach, I have no problem to implement it.
> It may be that the codepath that reads from an existing bitmap file
> says "ah, the file given to us seems to be in format X (either EWAH
> or ROARING or perhaps something else), so let's call bitmap_init(X)
> to obtain the in-core data structure to deal with that file". When
> that happens, you may probably need to have two cases in the default:
> arm of this switch statement, i.e. one to diagnose a BUG() to pass
> an uninitialized bitmap type to the codepath, and the other to
> diagnose a runtime error() to have read a bitmap file whose format
> this version of Git does not understand.
Ok, understood. Thanks.
> These repetitive patterns makes me wonder if void *bitmap
> is a good type to be passing around. Shouldn't it be a struct with
> its first member being a bitmap_type, and another member being what
> these functions are passing to the underlying bitmap format specific
> functions as "bitmap"? E.g.
>
> void bitmap_unset(struct bitmap *bm, uint32_t i)
> {
> switch (bm->type) {
> case EWAH:
> ewah_bitmap_remove(bm->u.ewah, i);
> break;
> ...
Good idea! Thanks.
> > +
> > +enum bitmap_type {
> > + INIT_BITMAP_TYPE = 0,
>
> "UNINITIALIZED_BITMAP_TYPE", probably.
Ok.
> > +void *roaring_or_ewah_bitmap_init(void);
>
> I would strongly suggest reconsider these names. What if you later
> want to add the third variant? roaring_or_ewah_or_xyzzy_bitmap_init()?
>
> Instead just use the most generic name, like "bitmap_init", perhaps
> something along the lines of ...
>
> struct bitmap {
> enum bitmap_type type;
> union {
> struct ewah_bitmap *ewah;
> struct roaring_bitmap *roaring;
> } u;
> };
>
> struct bitmap *bitmap_new(enum bitmap_type type)
> {
> struct bitmap *bm = xmalloc(sizeof(*bm));
>
> bm->type = type;
> switch (bm->type) {
> case EWAH:
> bm->u.ewah = ewah_new();
> break;
> case ROARING:
> bm->u.roaring = roaring_bitmap_create();
> break;
> default:
> die(_("unknown bitmap type %d"), (int)type);
> }
> return bm;
> }
Got it. It seems a better option than the current one.
Thanks )
next prev parent reply other threads:[~2022-09-30 16:24 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-19 17:47 [PATCH 0/5] [RFC] introduce Roaring bitmaps to Git Abhradeep Chakraborty via GitGitGadget
2022-09-19 17:47 ` [PATCH 1/5] reachability-bitmaps: add CRoaring library " Abhradeep Chakraborty via GitGitGadget
2022-09-19 17:47 ` [PATCH 2/5] roaring.[ch]: apply Git specific changes to the roaring API Abhradeep Chakraborty via GitGitGadget
2022-09-19 18:33 ` Derrick Stolee
2022-09-19 22:02 ` Junio C Hamano
2022-09-20 12:19 ` Derrick Stolee
2022-09-20 15:09 ` Abhradeep Chakraborty
2022-09-21 16:58 ` Junio C Hamano
2022-09-20 14:46 ` Abhradeep Chakraborty
2022-09-19 17:47 ` [PATCH 3/5] roaring: teach Git to write roaring bitmaps Abhradeep Chakraborty via GitGitGadget
2022-09-30 6:20 ` Junio C Hamano
2022-09-30 16:23 ` Abhradeep Chakraborty [this message]
2022-10-30 6:35 ` Abhradeep Chakraborty
2022-10-30 19:46 ` Derrick Stolee
2022-10-31 14:30 ` Abhradeep Chakraborty
2022-10-31 16:06 ` Junio C Hamano
2022-10-31 17:51 ` C99 -> C11 or C17? (was: [PATCH 3/5] roaring: teach Git to write roaring bitmaps) Ævar Arnfjörð Bjarmason
2022-10-31 20:55 ` rsbecker
2022-11-01 6:58 ` [PATCH 3/5] roaring: teach Git to write roaring bitmaps Abhradeep Chakraborty
2022-09-19 17:47 ` [PATCH 4/5] roaring: introduce a new config option for " Abhradeep Chakraborty via GitGitGadget
2022-09-19 17:47 ` [PATCH 5/5] roaring: teach Git to read " Abhradeep Chakraborty via GitGitGadget
2022-09-19 18:18 ` [PATCH 0/5] [RFC] introduce Roaring bitmaps to Git Derrick Stolee
2022-09-20 14:05 ` Abhradeep Chakraborty
2022-09-20 21:59 ` Taylor Blau
2022-09-21 15:27 ` Abhradeep Chakraborty
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: http://vger.kernel.org/majordomo-info.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAPOJW5yxRETdVk014gQYFud9_Nrt+OQGSVNQ8Pw2wDEMMFMm1Q@mail.gmail.com \
--to=chakrabortyabhradeep79@gmail.com \
--cc=derrickstolee@github.com \
--cc=git@vger.kernel.org \
--cc=gitgitgadget@gmail.com \
--cc=gitster@pobox.com \
--cc=kaartic.sivaraam@gmail.com \
--cc=me@ttaylorr.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/mirrors/git.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).