Re: SQL99 ARRAY support proposal - Mailing list pgsql-hackers

On Fri, 14 Mar 2003, Þórhallur Hálfdánarson wrote:

> -*- Greg Stark <gsstark@mit.edu> [ 2003-03-14 17:43 ]:
> > Do you really think someone looking for a function to break up a string into a
> > list of strings would ever think of looking up "explode" in an index if he
> > hadn't already used PHP or (shudder) VBScript?
> 
> If one had gotten used to Lotus Notes, sure. ;>

To try and get back on track...

Let me ask you, if you were looking through a list of array functions 
and you saw explode and implode, and you had no other experience with a 
language that used those keywords, would you, upon seeing them, have some 
idea what they did?

I certainly did the first time I saw them way back when on Rex.  Since 
there was no one around to teach me the religious belief that explode and 
implode are anathema to "real" programmers, I started using them and was 
quite happy with their performance.  (Sorry, that's a reference to a 
previouis post about "real" code.  I'll stop the verbal sparring now :-)

I would like to have a single word.  Join and Split are a poor choice 
becuase join already has a very well defined meaning in SQL.  any 
variation that keeps the word join in it is a poor choice because it can 
be confusing.  What other languages have used before is of little 
consequence really, as long as Postgresql's choices are internally 
consistent, are a desriptive word and don't conflict with current key 
words.

The fact that some languages that use explode / implode are not favored by 
certain folks means less than nothing to me.  The names for functions 
should be something that is easily found in the docs both when you don't 
know what it is called and when you do.

Explode may be a little harder to find when you don't know the name yet, 
but since you should be using the table of contents, not the index to look 
up a function whose name you don't know, then all that means is having a 
good table of contents and well layed out docs.

If you do know the function name, you shouldn't have 200 choices in an 
index or a keyword search.  Join, both as a keyword, and as a word, is 
very common in the docs.  Searching for it, I currently find 412 
entries for the word join in the 7.3.2 docs.

Explode, on the other hand, returns 0 count.  So, if someone goes to 
google and searches on site:techdocs.postgresql.org join, he will get 
hundreds of pages back.  If he searches on explode, he would, in the 
future, only find those pages referencing the explode function.

I'll repeat, the part about looking in an index for a keyword you don't 
know yet is a straw man, and a poor one at that.  If you don't know the 
keyword for a cursor is "declare" why would you look in the index?  You 
wouldn't, you'd look up cursors in the table of contents and find it that 
way.  Or you'd search for it.  If cursors were used for some other 
function in postgresql (widget counting et. al.) it would make it harder 
to find the entries you wanted for real cursors.