Re: PostgreSQL Developer Best Practices
От | Gavin Flower |
---|---|
Тема | Re: PostgreSQL Developer Best Practices |
Дата | |
Msg-id | 55DC3EE8.4060101@archidevsys.co.nz обсуждение исходный текст |
Ответ на | Re: PostgreSQL Developer Best Practices (Karsten Hilbert <Karsten.Hilbert@gmx.net>) |
Список | pgsql-general |
On 25/08/15 19:04, Karsten Hilbert wrote: > On Tue, Aug 25, 2015 at 02:02:17PM +1200, Gavin Flower wrote: > >> On 25/08/15 01:15, Ray Cote wrote: >>> On Sat, Aug 22, 2015 at 11:46 AM, Karsten Hilbert <Karsten.Hilbert@gmx.net >>> <mailto:Karsten.Hilbert@gmx.net>> wrote: >>> >> [...] >>> 9. Do NOT arbitrarily assign an "id" column to a table as a >>> primary key when other columns >>> are perfectly suited as a unique primary key. >>> >>> ... >>> >>> Good example: >>> CREATE TABLE accounts >>> ( accout_id bigint NOT NULL , >>> >>> >>> I would not consider the general use of natural primary keys to be best >>> practice. > Gavin, Ray, > > I certainly didn't write any of the above. > > Karsten Hi Karsten, It took me a couple of minutes, but I traced "9. ..." to melvin6925@gmail.com who opened the thread Looks like Ray misquoted back in the entry that can be identified by (using the 'source' option on my mail client) From: Ray Cote <rgacote@appropriatesolutions.com> Date: Mon, 24 Aug 2015 09:15:27 -0400 Message-ID: <CAG5tnzqTausEhFtRpfWCunx4YNFuGTFyUZyTkn5f2E7RaYKE=g@mail.gmail.com> which was On Sat, Aug 22, 2015 at 11:46 AM, Karsten Hilbert <Karsten.Hilbert@gmx.net> wrote: > > 1. Prefix ALL literals with an Escape > > EG: SELECT E'This is a \'quoted literal \''; > > SELECT E'This is an unquoted literal'; > > > > Doing so will prevent the annoying "WARNING: nonstandard use of > escape in a string literal" > I'd be concerned that what is missing here is the bigger issue of Best Practice #0: Use Bound Variables. The only way I've seen invalid literals show up in SQL queries is through the dynamic generation of SQL Statements vs. using bound variables. Not using bound variables is your doorway to SQL injection exploits. 9. Do NOT arbitrarily assign an "id" column to a table as a primary key > when other columns > are perfectly suited as a unique primary key. ... Good example: > CREATE TABLE accounts > ( accout_id bigint NOT NULL , I would not consider the general use of natural primary keys to be best practice. Let's assume your account_id field is used as a foreign key in a dozen other tables. 1) What happens if someone mis-types the account-id? To correct that, you also need to correct the FK field in the other dozen tables. 2) What happens when your company starts a new project (or buys a competitor) and all the new account numbers are alpha-numeric? 3) Your example shows the id as a bigint, but your rule is not limited to integers. What if your table is country populations and the primary key is country name? Now, you have quite large foreign keys (and a country changing its name is not unheard of). (and let's not even get started on case-sensitivity or character encodings). Cheers, Gavin
В списке pgsql-general по дате отправления: