[sf-perl] Some well-formed XML?

Andrew Sigmund ASIGMUND at altera.com
Thu Sep 22 08:38:10 PDT 2005

Every value for every NULLable column consists of two parts: A flag
indicating whether or not a value is present, and the value itself.
Although this seems like overkill when you actually have values (if you
have a value, then it's obviously there (not NULL)), it is just the
thing for distinguishing between absent values (NULLs) and "empty"

Alternatively, I suppose you could simply not include anything about the
column in rows in which the value is null.

ColumnB is not null and not empty:

ColumnB is not null, but is an empty string:

ColumnB is null:

-----Original Message-----
From: sanfrancisco-pm-bounces at pm.org
[mailto:sanfrancisco-pm-bounces at pm.org] On Behalf Of Joseph Brenner
Sent: Tuesday, September 20, 2005 15:57
To: San Francisco Perl Mongers User Group
Subject: Re: [sf-perl] Some well-formed XML?

Andrew Sigmund <ASIGMUND at altera.com> wrote:

> Joseph Brenner wrote: 
> > 
> > Rich Morin <rdm at cfcl.com> wrote:
> > 
> > >      So, it's fairly straightforward to serialize a set of RDBMS
> > >      tables as XML and load them back into another database.  The
> > >      rules for the (de-)serialization, however, will probably be
> > >      hard-coded into your dumping and loading code.
> > 
> > Funny, I was just in the process of doing just this sort of thing.
> > 
> > In particular I'm noodling around how to preserve NULLs. 
> > 
> > NULLs can be mapped to undefs in perl, which usually get mapped to
> > empty strings when written to some text format, but then that leaves
> > you with no way to distinguish between those and fields that started
> > out as empty (but non-NULL) VARCHARs.
> > 
> > So, you hack in code to translate the NULL/undefs to a string
> > Perhaps with additional hackery to preserve any odd ball strings 
> > containing "NULL" to start with? 

> It might help to think of NULLs as the DBMS thinks of them, that is,
> as values themselves, but as the absence of a value in a column that
> doesn't require a value.  NULLability is an attribute of the column,
> a column of any data-type (VARCHAR, INT, DATETIME, etc.) may have this
> attribute, and therefore not have a value, that is, "have a NULL

But how exactly does that help?  My problem is convincing the filesystem

to think of them the same way the DB does.

What I want to do is to perfectly preserve some table state by dumping
the contents to files.  I want to be able to run tests on some existing
code, then reset the tables the code uses to a known state, and then
re-run the tests.

My presumption is that somewhere in the code something may care that 
a field is NULL rather than empty: that's why I'm worrying about 
preserving NULLs.

SanFrancisco-pm mailing list
SanFrancisco-pm at pm.org

More information about the SanFrancisco-pm mailing list