It's certainly a worthy goal:

...to build towards a library of useful and hopefully generic sql statements.

Sadly, SQL doesn't admit genericity easily. SQL isn't a procedural language; that is both its strength and weakness. It doesn't admit variables (not "pure" SQL), so, for example, you can't create a "generic count-duplicates" statement like this:

SELECT @columns, COUNT(*)
FROM $table
GROUP BY @columns
HAVING COUNT(*) > 1
[download]

where the above is a mythical SQL-Perl hybrid.

You can get halfway there. In the past, I've often resorted to managing many statements with a bit of Perl. For example, in one of my scripts to rebuild tables you'll find stuff like this:

@key_cols1 = qw(account_id group_id);
@key_cols2 = qw(account_id manager_id asset_id);
# Later...
$set1_sql = 'SELECT ' . join(',', @key_cols1)
   . 'FROM ' . $table1_name
   . 'WHERE ' . $table1_cond
# You get the idea
[download]

So, it's not pretty, but it's better to maintain one set of templates in code than 18 separate SQL files.

You're right, it seems like there should be a better way, but most SQL interpreters seem to stop just a hair short or supporting cool stuff like this. If you can come up with something workable, I'll be the first to grab your stuff!