Date: Sun, 12 Mar 2023 17:06:03 -0500
From: Justin Pryzby <pryzby@telsasoft.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Matthias van de Meent <boekewurm+postgres@gmail.com>,
	Ilya Gladyshev <ilya.v.gladyshev@gmail.com>,
	Alvaro Herrera <alvherre@alvh.no-ip.org>,
	Tomas Vondra <tomas.vondra@enterprisedb.com>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: Progress report of CREATE INDEX for nested partitioned tables
Message-ID: <ZA5My7SUI0/VnmCS@telsasoft.com>
References: <20230201155336.GU22427@telsasoft.com>
 <CAEze2WgvJOvHv0mp3LwDbF27g7e_HgiWvbWdGH8980S14VBPoQ@mail.gmail.com>
 <DF2FEE0B-8D65-4926-B2E1-79C84363DBAC@gmail.com>
 <CAEze2Wgj=Ktn_B4U1sRRnWwitQcB4gLkb1QR8627bGRfzW4QOA@mail.gmail.com>
 <20230202151807.GY22427@telsasoft.com>
 <20230208224049.GO1653@telsasoft.com>
 <20230217002651.GW1653@telsasoft.com>
 <1000008.1678480570@sss.pgh.pa.us>
 <ZA4xlW9VB+1fYgyo@telsasoft.com>
 <1605789.1678652046@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1605789.1678652046@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/ZA5My7SUI0/VnmCS%40telsasoft.com>
Precedence: bulk

On Sun, Mar 12, 2023 at 04:14:06PM -0400, Tom Lane wrote:
> Justin Pryzby <pryzby@telsasoft.com> writes:
> > On Fri, Mar 10, 2023 at 03:36:10PM -0500, Tom Lane wrote:
> >> I took a look through this.  It seems like basically a good solution,
> >> but the count_leaf_partitions() function is bothering me, for two
> >> reasons:
> 
> > ... find_all_inheritors() will also have been called by
> > ProcessUtilitySlow().  Maybe it's sufficient to mention that ?
> 
> Hm.  Could we get rid of count_leaf_partitions by doing the work in
> ProcessUtilitySlow?  Or at least passing that OID list forward instead
> of recomputing it?

count_leaf_partitions() is called in two places:

Once to get PROGRESS_CREATEIDX_PARTITIONS_TOTAL.  It'd be easy enough to
pass an integer total via IndexStmt (but I think we wanted to avoid
adding anything there, since it's not a part of the statement).

count_leaf_partitions() is also called for sub-partitions, in the case
that a matching "partitioned index" already exists, and the progress
report needs to be incremented by the number of leaves for which indexes
were ATTACHED.  We'd need a mapping from OID => npartitions (or to
compile some data structure of all the partitioned partitions).  I guess
CreateIndex() could call CreatePartitionDirectory().  But it looks like
that would be *more* expensive.

-- 
Justin