Re: [QUESTIONS] varchar vs text

Bruce Momjian wrote:

In v6.3, indices are supported with use of ~ and LIKE operators.
An unofficial patch for v6.2 is also available to do this.

Hopefully, the developers will have remembered to change the documentation
as well.

Do you have a suggestion on where to put such a mention?

This should be in release notes in the hardcopy/html documentation. I'd like to
work on integrating these soon after v6.3 is released, and then do updates
there. That way, there will be a place for people to keep track of significant
changes and improvements which would affect a user.

Much of the new docs which is release-specific is like to old docs. Ugly, but I
simply didn't have enough time to rewrite _all_ 200 pages :)

Varchar currently (in 6.2.1 and below) takes up the entire length specified
in the definition, despite the fact the value in it may actually be
shorter. Text takes only the space taken by the value.

Thanks for the clarification. In this case, what happens with varchar's
length if the original definition for that field leaves length undefined?
Does it behave like text in that case?

You really shouldn't be doing that. Not sure what happens. Not a good
idea:

create table test (x varchar);

?? This was defined to be a varchar of unlimited length, much like, or identical
to, text. Should this now be disallowed? If so, we can fix the parser to disallow
it so people don't get misled.

Oh, I didn't know. There really is no difference between varchar with
no lenght, and text, but if it doesn't break anything, no problem.

I also vaguely recall seeing a message last year about the use of indexes
in queries: that in [some circumstances] indexes built on varchar fields
don't get used and a sequential scan through all records takes place
instead. Is there any distinction between varchar and text here?

Don't remember that.

This was probably Bruce's improvements to allow indices on some pattern matching.
Doesn't make a distinction between these types in its behavior.

Not sure what to say on this. I remember that issue, but not how it
caused any problem.

-- 
Bruce Momjian                          |  830 Blythe Avenue
maillist@candle.pha.pa.us              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)

I had a chance to look at the users and programmers manuals you just
installed. Very nice. Lots of new stuff and cleanup, and you
integrated much of the separate documentation in one place. I have
added a mention of it in my release summary.

I can easily send you html of what I am doing. The FAQ is already html,
and the TODO list is ascii, but converted using txt2html from
http://www.cs.wustl.edu/~seth/txt2html/. Works really well. It
recoginizes certain text formatting styles, and outputs HTML to make it
look correct on a web page. Perhaps we could use that to convert over
some of the ASCII-only stuff we have.

Yes, that would help, and then I can run a brute-force filter to convert the html to
almost-DocBook sgml. From there on we can turn it around and generate html from the
DocBook sources, for posting on the web page etc.

Seems it may be nice to have all the docs in the separate directories
all in html, and have 'make' grab them and convert them into the manual.
I really don't know what is involved, or whether you can just grab html
and place it into sgml documents, but it is an idea. Actually, the
doc/src/*.sgml files look pretty easy to understand, so maybe we all
need to learn it.

Well, it works _almost_ like this. Without getting caught up in the fact that html
_is_ sgml, just not sufficient to fully specify document content, the document
source would all be in DocBook sgml, then converted to html, hardcopy, ascii, and
man pages from there.

DocBook has a learning curve when starting from scratch, but I've put in the 100
hours to get over that hump. From here on, the docs can evolve from existing
documents, and stealing formatting specs from those will make a new doc easy to
write.

For each of the current plain text, man page, or html _source_ docs we will need to
get the maintainer to agree to try using sgml for that. I'll do, or assist with, the
conversion to sgml and from then on the maintainer would make maintenance changes to
the sgml source. I figured we can tackle that one at a time over the next couple of
months.

I guess the manual is so nice, I want to make sure it can stay
up-to-date without much effort on your part. I am sure you have already
thought of that.

Well, that is the advantage to using sgml, as long as others are willing to maintain
information in that format. I'll stress that _new_ information can be written
without sgml in plain text and someone can then help convert it. From then on, it
would be easiest if it were maintained from the sgml sources.

You have certainly jump-started our documentation, and now that it is so
nice, I am sure people will start getting involved.

Thanks. I really hope so :)

Lots of open issues with content, presentation, etc. and as we discuss it on the
Docs list we can start a ToDo to keep track of where we are headed.

- Tom

I had a chance to look at the users and programmers manuals you just
installed. Very nice. Lots of new stuff and cleanup, and you
integrated much of the separate documentation in one place. I have
added a mention of it in my release summary.

I can easily send you html of what I am doing. The FAQ is already html,
and the TODO list is ascii, but converted using txt2html from
http://www.cs.wustl.edu/~seth/txt2html/. Works really well. It
recoginizes certain text formatting styles, and outputs HTML to make it
look correct on a web page. Perhaps we could use that to convert over
some of the ASCII-only stuff we have.

Yes, that would help, and then I can run a brute-force filter to convert the html to
almost-DocBook sgml. From there on we can turn it around and generate html from the
DocBook sources, for posting on the web page etc.

OK. I recommend you just grab the TODO and FAQ from the web site,
unless you want HTML versions of them in the distribution along with the
ASCII verions.

-- 
Bruce Momjian                          |  830 Blythe Avenue
maillist@candle.pha.pa.us              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)

I can easily send you html of what I am doing. The FAQ is already html,
and the TODO list is ascii, but converted using txt2html from
http://www.cs.wustl.edu/~seth/txt2html/. Works really well. It
recoginizes certain text formatting styles, and outputs HTML to make it
look correct on a web page. Perhaps we could use that to convert over
some of the ASCII-only stuff we have.

Yes, that would help, and then I can run a brute-force filter to convert the html to
almost-DocBook sgml. From there on we can turn it around and generate html from the
DocBook sources, for posting on the web page etc.

OK. I recommend you just grab the TODO and FAQ from the web site,
unless you want HTML versions of them in the distribution along with the
ASCII verions.

Assuming we aren't doing this until post-v6.3 release, will let you know when we are
ready to start the conversion. Need to draw a line at how much can go into v6.3, and I
think we are past it wrt the docs except for perhaps goof-up fixes of the packages.

My first project after v6.3 will be getting jade/DocBook going on postgresql.org (perhaps
it already is; Marc pointed me at something which looked like a jade package). Then, we
can demonstrate how to run it on that machine, and perhaps tie it in to an automatic html
documentation update from cron or from cvs. Also, I'm hoping to be busy answering
questions and helping all those new documenters out there :)

- Tom

On Sun, 1 Mar 1998, Thomas G. Lockhart wrote:

My first project after v6.3 will be getting jade/DocBook going on
postgresql.org (perhaps it already is; Marc pointed me at something
which looked like a jade package). Then, we can demonstrate how to run
it on that machine, and perhaps tie it in to an automatic html
documentation update from cron or from cvs. Also, I'm hoping to be busy
answering questions and helping all those new documenters out there :)

jade was installed ~Jan 13th :) Of course, it hasn't been tested
yet, but let me know if there are any problems :)

Marc G. Fournier
Systems Administrator @ hub.org
primary: scrappy@hub.org secondary: scrappy@{freebsd|postgresql}.org

jade was installed ~Jan 13th :) Of course, it hasn't been tested
yet, but let me know if there are any problems :)

OK, jade is there, but I need the DocBook DTD integrated into jade's catalog:

gmake admin.tar.gz

gmake -C sgml clean
gmake[1]: Entering directory `/home/users/t/thomas/pgsql/doc/src/sgml'
(rm -rf *.html *.htm)
gmake[1]: Leaving directory `/home/users/t/thomas/pgsql/doc/src/sgml'
gmake -C sgml admin.html
gmake[1]: Entering directory `/home/users/t/thomas/pgsql/doc/src/sgml'
(rm -rf *.htm)
jade -D sgml -d /home/users/t/thomas/db107.d/docbook/html/docbook.dsl -t sgml
admin.sgml
jade:admin.sgml:8:59:W: cannot generate system identifier for public text
"-//Davenport//DTD DocBook V3.0//EN"
jade:admin.sgml:19:0:E: reference to entity "BOOK" for which no system identifier
could be generated
jade:admin.sgml:8:0: entity was defined here
jade:admin.sgml:19:0:E: DTD did not contain element declaration for document type
name
jade:admin.sgml:21:5:E: element "BOOK" undefined
jade:admin.sgml:25:6:E: element "TITLE" undefined
...

I had given you a reference for the source packages for my installation; do you
need that again? I think, as a first step, we just need the catalog stuff
updated.

- Tom