Re: Typos/Questions in bloom documentation - Mailing list pgsql-hackers

From David G. Johnston
Subject Re: Typos/Questions in bloom documentation
Date
Msg-id CAKFQuwbczigM80D8njmk_vsSk6xKQtt6BH3C6Sad3M+G+oKtWA@mail.gmail.com
Whole thread Raw
In response to Re: Typos/Questions in bloom documentation  (Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>)
Responses Re: Typos/Questions in bloom documentation  (Michael Paquier <michael.paquier@gmail.com>)
List pgsql-hackers
On Wed, Apr 20, 2016 at 9:18 PM, Amit Langote <Langote_Amit_f8@lab.ntt.co.jp> wrote:
On 2016/04/21 11:19, David G. Johnston wrote:
> On Wednesday, April 20, 2016, Amit Langote wrote:
>> On 2016/04/21 6:51, David G. Johnston wrote:
>>> http://www.postgresql.org/docs/devel/static/bloom.html
>>>
>>> F.4.3 Examples
>>>
>>> Claims that the signature length is 80 bits - shouldn't it be 8?
>>
>> In F.4.1. Introduction:
>>
>> ... The user can specify signature length (in uint16, default is 5)
>>
>> So, it seems right to me.
>>
> Great.  Maybe you can consider re-wording it so others can understand.  I
> have no clue how 80bits is determined.  The phase you quote is obtuse to
> the casual user as well.  If that means 16x5=80 irrespective of columns it
> is not clear.

I agree it's unclear.  Does the following make it any better (updated
patch attached):

-   The user can specify signature length (in uint16, default is 5) and the
-   number of bits, which can be set per attribute (1 < colN < 2048).
+   The user can specify signature length in units of 16 bits (default is 5)
+   and the number of bits per indexed attribute.


​Better.  The "and" is confusing.  Is the signature length the sum of 16x5 + (bits per indexed attribute)​?

By the way, now I am slightly confused as well about per-column bits
assignment thing:

In F.4.1. Introduction:

... and the number of bits, which can be set per attribute (1 < colN < 2048).

And then in F.4.2. Parameters:

bloom indexes accept the following parameters in the WITH clause.

length
    Length of signature in uint16 type values

How about: "Number of 16bit units to use for the signature"


col1 — col16
    Number of bits for corresponding column

Which is it: col1 - col2048 or col1 - col16?  Or are they different things
altogether?

​Good question...

David J.​
 

pgsql-hackers by date:

Previous
From: "David G. Johnston"
Date:
Subject: Re: Query Procedures
Next
From: Kevin Grittner
Date:
Subject: Re: [COMMITTERS] pgsql: Inline initial comparisons in TestForOldSnapshot()