| From: | Robert Haas <robertmhaas(at)gmail(dot)com> |
|---|---|
| To: | "mac_man2005(at)hotmail(dot)it" <mac_man2005(at)hotmail(dot)it> |
| Cc: | pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: About tapes |
| Date: | 2010-06-18 19:00:15 |
| Message-ID: | AANLkTinERwYkDbw-EFuuUQo1R5YfhSiFIAQ_s_UwyRiB@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Fri, Jun 18, 2010 at 2:36 PM, mac_man2005(at)hotmail(dot)it
<mac_man2005(at)hotmail(dot)it> wrote:
> Please take a look at the initial comment contained into the logtape.c file:
> http://doxygen.postgresql.org/logtape_8c-source.html
>
> Almost at the beginning of that file, it is affirmed that implementing tapes
> on disk (quote: by creating a separate file for each "tape") will require
> more space than implementing merge on tapes themselves.
> Now, taking in account that tuplesort.c and logtape.c actually DO implement
> tapes on disk, in which case it would require between 2x and 4x the input
> space?
Did you read the rest of the comment? It explains how the code avoids this...
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company
| From | Date | Subject | |
|---|---|---|---|
| Next Message | mac_man2005@hotmail.it | 2010-06-18 19:11:57 | Re: About tapes |
| Previous Message | mac_man2005@hotmail.it | 2010-06-18 18:36:31 | About tapes |