Обсуждение: "an SQL" vs. "a SQL"

I thought it might be worth having this conversation before we branch for v15.

It seems we have no standard as to if we say "a SQL" or "an SQL".

Personally, I pronounce the language as es-que-ell, so I'd write "an
SQL".  If you say "sequel", then you'll think differently. The reason
I do this is that the language was only briefly named sequel but was
renamed to SQL. For me calling it sequel seems wrong or out-dated. End
of personal opinion.

Let this thread not become the place where you tell me why I'm wrong.
Let's just get some consensus on something, make a change then move
on.

Overall we seem to mostly write "a SQL".

~/pg_src$ git grep -E "\s(a|A)\sSQL\s" | wc -l
855
~/pg_src$ git grep -E "\s(an|An)\sSQL\s" | wc -l
295

However, we mostly use "an SQL"  in the docs.

~/pg_src$ cd doc/
~/pg_src/doc$ git grep -E "\s(a|A)\sSQL\s" | wc -l
55
~/pg_src/doc$ git grep -E "\s(an|An)\sSQL\s" | wc -l
94

I think we should change all 55 instances of "a SQL" in the docs to
use "an SQL" and leave the 800 other instances of "a SQL" alone.
Changing those does not seem worthwhile as it could cause
back-patching pain.

I mostly think that because of the fact that my personal opinion
agrees with the majority of instances in the docs. Makes more sense to
change 55 places than 94 places.

Interesting reading:
http://patorjk.com/blog/2012/01/26/pronouncing-sql-s-q-l-or-sequel/

Further, there might be a few more in the docs that we might want to
consider changing:

git grep -E "\sa\s(A|E|F|H|I|L|M|N|O|S|X)[A-Z]{2,5}\s"

I see "a FSM", "a FIFO", "a SSPI", "a SASL", "a MCV", "a SHA", "a SQLDA"

My regex foo is not strong enough to think how I might find multiline instances.

David

On 10.06.21 09:26, David Rowley wrote:
> It seems we have no standard as to if we say "a SQL" or "an SQL".

The SQL standard uses "an SQL-something".

> However, we mostly use "an SQL"  in the docs.
> 
> ~/pg_src$ cd doc/
> ~/pg_src/doc$ git grep -E "\s(a|A)\sSQL\s" | wc -l
> 55
> ~/pg_src/doc$ git grep -E "\s(an|An)\sSQL\s" | wc -l
> 94
> 
> I think we should change all 55 instances of "a SQL" in the docs to
> use "an SQL" and leave the 800 other instances of "a SQL" alone.
> Changing those does not seem worthwhile as it could cause
> back-patching pain.

agreed

> Further, there might be a few more in the docs that we might want to
> consider changing:
> 
> git grep -E "\sa\s(A|E|F|H|I|L|M|N|O|S|X)[A-Z]{2,5}\s"
> 
> I see "a FSM", "a FIFO", "a SSPI", "a SASL", "a MCV", "a SHA", "a SQLDA"
> 
> My regex foo is not strong enough to think how I might find multiline instances.

Um, of those, I pronounce FIFO, SASL, and SHA as words, with an "a" article.

> On 10 Jun 2021, at 10:54, Dave Page <dpage@pgadmin.org> wrote:

> .. I would perhaps suggest extending to any user-visible messages in the code.

I agree, consistent language between docs and user-facing messages is
important.

--
Daniel Gustafsson        https://vmware.com/

On Thu, 10 Jun 2021 at 20:58, Daniel Gustafsson <daniel@yesql.se> wrote:
>
> > On 10 Jun 2021, at 10:54, Dave Page <dpage@pgadmin.org> wrote:
>
> > .. I would perhaps suggest extending to any user-visible messages in the code.
>
> I agree, consistent language between docs and user-facing messages is
> important.

Yeah, agreed.

I came up with the attached patch.

The only additional abbreviation that I found to be incorrect that I'd
previously not mentioned was "SRF".  I changed that to use "an".

I only found 4 error messages that needed to be updated.  There's some
incorrect stuff remaining in a few README files which I couldn't
decide if I should update or not.

Most of the offenders away from the docs are the translator hint
comments and within the .po files themselves.

$ git grep -E "translator:.*(a|A)\sSQL" | wc -l
690

Only 816 instances of "a SQL" remain, so only 126 are not related to
translator hints or .po files.

Does anyone have any thoughts if the READMEs should be fixed up?

David

On 2021-Jun-10, David Rowley wrote:

> I thought it might be worth having this conversation before we branch for v15.
> 
> It seems we have no standard as to if we say "a SQL" or "an SQL".

I was just reading the standard a couple of days ago and happened to
notice that the standard itself in some places uses "a SQL" and in other
places "an SQL".  I didn't stop to make an analysis of that, so I don't
know how prevalent each form is -- I just giggled and moved on.

> My regex foo is not strong enough to think how I might find multiline instances.

This catches some of these:

ag "\sa[\s*]*\n[\s*]*(A|E|F|H|I|L|M|N|O|S|X)[A-Z]{2,5}\s"

You get a bunch of "a NULL" or "a NOT" and so on, but here's a few valid ones:

contrib/tablefunc/tablefunc.c:316: * crosstab - create a crosstab of rowids and values columns from a
contrib/tablefunc/tablefunc.c:317: * SQL statement returning one rowid column, one category column,

contrib/tablefunc/tablefunc.c:607: * crosstab - create a crosstab of rowids and values columns from a
contrib/tablefunc/tablefunc.c:608: * SQL statement returning one rowid column, one category column,

doc/src/sgml/plpgsql.sgml
1127:     The result of a
1128:     SQL command yielding a single row (possibly of multiple

src/backend/catalog/pg_subscription.c:438:             * translator: first %s is a SQL ALTER command and second %s is
a
src/backend/catalog/pg_subscription.c:439:             * SQL DROP command

src/backend/replication/logical/logical.c:126:     * 1) We need to be able to correctly and quickly identify the
timelinea
 
src/backend/replication/logical/logical.c:127:     *      LSN belongs to

src/backend/libpq/auth.c:847:     * has.  If it's an MD5 hash, we must do MD5 authentication, and if it's a
src/backend/libpq/auth.c:848:     * SCRAM secret, we must do SCRAM authentication.

-- 
Álvaro Herrera       Valdivia, Chile

On Fri, 11 Jun 2021 at 02:04, David Rowley <dgrowleyml@gmail.com> wrote:
> I came up with the attached patch.

Further searching using:

git grep -E "\s(an|An)\s(F|H|L|M|N|S|X)[A-Z]{2,5}"

(i.e vowel sounding, but not actually starting with a vowel then
manually looking for pronounceable ones.)

-  by a response from client in an SASLResponse message. The particulars of
+  by a response from client in a SASLResponse message. The particulars of

-     An SHA1 hash of the random prefix and data is appended.
+     A SHA1 hash of the random prefix and data is appended.

-       requires an MIT Kerberos installation and opens TCP/IP listen sockets.
+       requires a MIT Kerberos installation and opens TCP/IP listen sockets.

I think all of these should use "a" rather than "an".

David

On Fri, 11 Jun 2021 at 02:35, Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:
>
> On 2021-Jun-10, David Rowley wrote:
> > My regex foo is not strong enough to think how I might find multiline instances.
>
> This catches some of these:
>
> ag "\sa[\s*]*\n[\s*]*(A|E|F|H|I|L|M|N|O|S|X)[A-Z]{2,5}\s"

Thanks. I ended up using -C 1 and manually checking the previous line.

> You get a bunch of "a NULL" or "a NOT" and so on, but here's a few valid ones:
>
> contrib/tablefunc/tablefunc.c:316: * crosstab - create a crosstab of rowids and values columns from a
> contrib/tablefunc/tablefunc.c:317: * SQL statement returning one rowid column, one category column,
>
> contrib/tablefunc/tablefunc.c:607: * crosstab - create a crosstab of rowids and values columns from a
> contrib/tablefunc/tablefunc.c:608: * SQL statement returning one rowid column, one category column,
>
> doc/src/sgml/plpgsql.sgml
> 1127:     The result of a
> 1128:     SQL command yielding a single row (possibly of multiple
>
> src/backend/catalog/pg_subscription.c:438:                       * translator: first %s is a SQL ALTER command and
second%s is a
 
> src/backend/catalog/pg_subscription.c:439:                       * SQL DROP command
>
> src/backend/replication/logical/logical.c:126:   * 1) We need to be able to correctly and quickly identify the
timelinea
 
> src/backend/replication/logical/logical.c:127:   *        LSN belongs to
>
> src/backend/libpq/auth.c:847:    * has.  If it's an MD5 hash, we must do MD5 authentication, and if it's a
> src/backend/libpq/auth.c:848:    * SCRAM secret, we must do SCRAM authentication.

Thanks. I've left all the .c file comments alone for no and looks like
I got the doc/src/sgml/plpgsql.sgml one already.

David

Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> On 2021-Jun-10, David Rowley wrote:
>> It seems we have no standard as to if we say "a SQL" or "an SQL".

> I was just reading the standard a couple of days ago and happened to
> notice that the standard itself in some places uses "a SQL" and in other
> places "an SQL".  I didn't stop to make an analysis of that, so I don't
> know how prevalent each form is -- I just giggled and moved on.

Indeed.  I think this is entirely pointless; there's zero hope that
any consistency you might establish right now will persist very long.
The largest effect of this proposed patch will be to create
back-patching headaches.

            regards, tom lane

On Fri, 11 Jun 2021 at 02:53, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Indeed.  I think this is entirely pointless; there's zero hope that
> any consistency you might establish right now will persist very long.
> The largest effect of this proposed patch will be to create
> back-patching headaches.

hmm.  Yet we do have other standards which we do manage to maintain.

I did limit the scope to just the docs and error messages. My thoughts
were that someone fudging a backpatch on the docs seems less likely to
cause a nuclear meltdown than someone doing the same in .c code.

David

David Rowley <dgrowleyml@gmail.com> writes:
> On Fri, 11 Jun 2021 at 02:53, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Indeed.  I think this is entirely pointless; there's zero hope that
>> any consistency you might establish right now will persist very long.

> hmm.  Yet we do have other standards which we do manage to maintain.

If there were some semblance of an overall consensus on the spelling,
I'd be fine with weeding out the stragglers.  But when the existing
usages are only about 2-to-1 in one direction or the other, I feel
quite confident in predicting that incoming patches are often going
to get this wrong.  Especially so if the convention you want to
establish in the docs is contrary to the majority usage in the code
comments --- how is that not going to confuse people?

            regards, tom lane

On Fri, 11 Jun 2021 at 03:24, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> If there were some semblance of an overall consensus on the spelling,
> I'd be fine with weeding out the stragglers.  But when the existing
> usages are only about 2-to-1 in one direction or the other, I feel
> quite confident in predicting that incoming patches are often going
> to get this wrong.

I'm pretty sure you're right and we will get some inconsistencies
creeping back in. I'm not really sure why you think that will be hard
to fix though.  If we catch them soon enough then we won't need to
worry about causing future backpatching pain.

> Especially so if the convention you want to
> establish in the docs is contrary to the majority usage in the code
> comments --- how is that not going to confuse people?

Why would someone go and gawk at code comments to clear up their
confusion about what they should write in the docs?  I think any sane
person that's looking for inspiration would look at the docs first.

I really think it's worth the trouble here to be consistent in our
public-facing documents.  When I read [1] earlier and the blog started
talking about Oracle documentation using sequel consistently before
going on to talk about MySQL's documentation, I started to get a bit
worried that the author might mention something about our lack of
consistency. I was glad to see they missed us out of that.  However,
maybe that's because we are inconsistent.

If you really feel that strongly about not changing this then I can
drop this. However, I'll likely growl every time I see "a SQL" in the
docs from now on.

David

[1] http://patorjk.com/blog/2012/01/26/pronouncing-sql-s-q-l-or-sequel/

David Rowley <dgrowleyml@gmail.com> writes:
> If you really feel that strongly about not changing this then I can
> drop this. However, I'll likely growl every time I see "a SQL" in the
> docs from now on.

[ shrug... ]  I'm not going to stand in your way.  However, I'm also
unlikely to worry about this point when copy-editing docs.

            regards, tom lane

On 11/06/21 2:48 am, Isaac Morland wrote:
> On Thu, 10 Jun 2021 at 10:43, David Rowley <dgrowleyml@gmail.com 
> <mailto:dgrowleyml@gmail.com>> wrote:
>
>     -      requires an MIT Kerberos installation and opens TCP/IP
>     listen sockets.
>     +       requires a MIT Kerberos installation and opens TCP/IP
>     listen sockets.
>
>     I think all of these should use "a" rather than "an".
>
>
> “A MIT …”? As far as I know it is pronounced M - I - T, which would 
> imply that it should use “an”. The following page seems believable and 
> is pretty unequivocal on the issue:
>
> https://mitadmissions.org/blogs/entry/como_se_dice/ 
> <https://mitadmissions.org/blogs/entry/como_se_dice/>
>
The rule is, in English, is that if the word sounds like it starts with 
a vowel then use 'an' rather than 'a'.  Though some people think that 
the rule only applies to words beginning with a vowel, which is a 
misunderstanding.

So 'an SQL' and 'an MIT'  are correct.   IMHO


Cheers,
Gavin

On 11/06/21 8:17 am, Isaac Morland wrote:
> On Thu, 10 Jun 2021 at 16:11, Gavin Flower 
> <GavinFlower@archidevsys.co.nz <mailto:GavinFlower@archidevsys.co.nz>> 
> wrote:
>
>     On 11/06/21 2:48 am, Isaac Morland wrote:
>
>     > “A MIT …”? As far as I know it is pronounced M - I - T, which would
>     > imply that it should use “an”. The following page seems
>     believable and
>     > is pretty unequivocal on the issue:
>     >
>     > https://mitadmissions.org/blogs/entry/como_se_dice/
>     <https://mitadmissions.org/blogs/entry/como_se_dice/>
>     > <https://mitadmissions.org/blogs/entry/como_se_dice/
>     <https://mitadmissions.org/blogs/entry/como_se_dice/>>
>     >
>     The rule is, in English, is that if the word sounds like it starts
>     with
>     a vowel then use 'an' rather than 'a'.  Though some people think that
>     the rule only applies to words beginning with a vowel, which is a
>     misunderstanding.
>
>     So 'an SQL' and 'an MIT'  are correct.   IMHO
>
>
> Right, spelling is irrelevant, it's about whether the word begins with 
> a vowel *sound*. Or so I've always understood and I'm pretty sure if 
> you listen to what people actually say that's what you'll generally 
> hear. So "A uranium mine" not "An uranium mine" since "uranium" begins 
> with a "y-" sound just like "yesterday". The fact that "u" is a vowel 
> is irrelevant. But then there is "an historic occasion" so go figure.
>
The 'h' in 'historic' is silent, at least it used to be -- I think now 
it is almost silent.  So using 'an historic occasion' is correct.

Gavin Flower <GavinFlower@archidevsys.co.nz> writes:
> On 11/06/21 8:17 am, Isaac Morland wrote:
>> ... But then there is "an historic occasion" so go figure.

> The 'h' in 'historic' is silent, at least it used to be -- I think now 
> it is almost silent.  So using 'an historic occasion' is correct.

It's silent according to the Brits, I believe.  In America, the
pronunciation varies.

            regards, tom lane

On 6/10/21 5:32 PM, Tom Lane wrote:
> Gavin Flower <GavinFlower@archidevsys.co.nz> writes:
>> On 11/06/21 8:17 am, Isaac Morland wrote:
>>> ... But then there is "an historic occasion" so go figure.
>> The 'h' in 'historic' is silent, at least it used to be -- I think now 
>> it is almost silent.  So using 'an historic occasion' is correct.
> It's silent according to the Brits, I believe.  In America, the
> pronunciation varies.
>
>             


I suspect "an historic" is bordering on archaic even in the UK these days.


cheers


andrew


--
Andrew Dunstan
EDB: https://www.enterprisedb.com

On Fri, 11 Jun 2021 at 02:48, Isaac Morland <isaac.morland@gmail.com> wrote:
>
> On Thu, 10 Jun 2021 at 10:43, David Rowley <dgrowleyml@gmail.com> wrote:
>>
>> -       requires an MIT Kerberos installation and opens TCP/IP listen sockets.
>> +       requires a MIT Kerberos installation and opens TCP/IP listen sockets.
>>
>> I think all of these should use "a" rather than "an".
>
>
> “A MIT …”? As far as I know it is pronounced M - I - T, which would imply that it should use “an”. The following page
seemsbelievable and is pretty unequivocal on the issue: 
>
> https://mitadmissions.org/blogs/entry/como_se_dice/

Opps. I'm not sure what I was thinking there.  I'd just been listening
to something in German, so maybe I'd had the German word in mind
instead.

David

On Fri, 11 Jun 2021 at 09:39, Andrew Dunstan <andrew@dunslane.net> wrote:
> I suspect "an historic" is bordering on archaic even in the UK these days.

Yeah, that's a weird one. Maybe
https://en.wikipedia.org/wiki/H-dropping is to blame.

David

On Fri, 11 Jun 2021 at 04:04, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> However, I'm also
> unlikely to worry about this point when copy-editing docs.

I'm sorry to hear that.  Maybe keeping this consistent will be one of
those endless jobs like keeping the source code pgindented. We still
try to keep that in order despite the audience for the source code
being much smaller than the audience for our documents.

Anyway, I'll set an alarm for this time next year so I can check on
how many inconsistencies have crept back in over the development
cycle.

In the meantime, I've pushed the fixes to master.

David

On Thu, Jun 10, 2021 at 05:39:00PM -0400, Andrew Dunstan wrote:
> I suspect "an historic" is bordering on archaic even in the UK these days.

Don't trigger me on the difference between "historic" and "historical"!  ;-)

(Hey, not every day I get to trim quoted text to one line --- see recent
pgsql-general discussion of the topic.)

-- 
  Bruce Momjian  <bruce@momjian.us>        https://momjian.us
  EDB                                      https://enterprisedb.com

  If only the physical world exists, free will is an illusion.

On Sun, Jun 13, 2021 at 07:36:54AM +0100, Geoff Winkless wrote:
> On Thu, 10 Jun 2021, 15:35 Alvaro Herrera, <alvherre@alvh.no-ip.org> wrote:
>> src/backend/libpq/auth.c:847:    * has.  If it's an MD5 hash, we must do
>> MD5 authentication, and if it's a
>> src/backend/libpq/auth.c:848:    * SCRAM secret, we must do SCRAM
>> authentication.
>
> Not sure whether you were just listing examples and you weren't suggesting
> this should be changed, but surely "SCRAM" is pronounced "scram" and is
> thus "a SCRAM"?

RFC 5802 uses "a SCRAM something" commonly, but "a SCRAM" alone does
not make sense:
https://datatracker.ietf.org/doc/html/rfc5802

The sentences quoted above look fine to me.
--
Michael

On 6/13/21 7:13 AM, Michael Paquier wrote:
> On Sun, Jun 13, 2021 at 07:36:54AM +0100, Geoff Winkless wrote:
>> On Thu, 10 Jun 2021, 15:35 Alvaro Herrera, <alvherre@alvh.no-ip.org> wrote:
>>> src/backend/libpq/auth.c:847:    * has.  If it's an MD5 hash, we must do
>>> MD5 authentication, and if it's a
>>> src/backend/libpq/auth.c:848:    * SCRAM secret, we must do SCRAM
>>> authentication.
>> Not sure whether you were just listing examples and you weren't suggesting
>> this should be changed, but surely "SCRAM" is pronounced "scram" and is
>> thus "a SCRAM"?
> RFC 5802 uses "a SCRAM something" commonly, but "a SCRAM" alone does
> not make sense:
> https://datatracker.ietf.org/doc/html/rfc5802
>
> The sentences quoted above look fine to me.


I don't think anyone was suggesting SCRAM should be used as a noun
rather than as an adjective. But adjectives can be preceded by an
indefinite article just as nouns can. The discussion simply left out the
implied following noun.


cheers


andrew


--
Andrew Dunstan
EDB: https://www.enterprisedb.com

On Fri, 11 Jun 2021 at 13:44, David Rowley <dgrowleyml@gmail.com> wrote:
> Anyway, I'll set an alarm for this time next year so I can check on
> how many inconsistencies have crept back in over the development
> cycle.

That alarm went off today.

There seem to be only 3 "a SQL"s in the docs to change to "an SQL".

This is a pretty old thread, so here's a link [1] to the discussion.

David

[1] https://postgr.es/m/CAApHDvpML27UqFXnrYO1MJddsKVMQoiZisPvsAGhKE_tsKXquw@mail.gmail.com

On Tue, Apr 11, 2023 at 05:43:04PM +1200, David Rowley wrote:
> That alarm went off today.
>
> There seem to be only 3 "a SQL"s in the docs to change to "an SQL".
>
> This is a pretty old thread, so here's a link [1] to the discussion.

Good catches!
--
Michael

On Tue, 11 Apr 2023 at 17:43, David Rowley <dgrowleyml@gmail.com> wrote:
>
> On Fri, 11 Jun 2021 at 13:44, David Rowley <dgrowleyml@gmail.com> wrote:
> > Anyway, I'll set an alarm for this time next year so I can check on
> > how many inconsistencies have crept back in over the development
> > cycle.
>
> That alarm went off today.
>
> There seem to be only 3 "a SQL"s in the docs to change to "an SQL".
>
> This is a pretty old thread, so here's a link [1] to the discussion.
>
> [1] https://postgr.es/m/CAApHDvpML27UqFXnrYO1MJddsKVMQoiZisPvsAGhKE_tsKXquw@mail.gmail.com

Link to the old thread above.

There's just 1 instance of "a SQL" that crept into PG16 after
d866f0374.  This probably means I'd be better off doing this in June a
few weeks before branching...

Patch attached.

David

On Tue, 9 Apr 2024 at 16:18, David Rowley <dgrowleyml@gmail.com> wrote:
> There's just 1 instance of "a SQL" that crept into PG16 after
> d866f0374.  This probably means I'd be better off doing this in June a
> few weeks before branching...
>
> Patch attached.

Pushed.

David