Re: Unmodifiable ResultSet wrapper?
On Friday, September 2, 2011 7:30:24 AM UTC-7, Ian Pilcher wrote:
On 08/29/2011 04:08 PM, Lew wrote:
Ian Pilcher wrote:
Is there any standard way to wrap a ResultSet into an unmodifiable
wrapper? I need to pass a ResultSet to a callback as I iterate throug=
h
it, and the callback has no business altering its state (calling next(=
),
For the sake of posterity, I ended up using a dynamic proxy. The
invocation handler matches the name of the method being called against
a whitelist (a HashSet of method names) and throws an
UnsupportedOperationException if the method name isn't in the whitelist.
The only fly in the ointment is the need to handle getMetaData()
specially and return a ResultSetMetaData proxy which whitelists every
method except unwrap().
There is no standard way to pass an unmodifiable ResultSet back, perhap=
s because it's a really bad idea. ResultSets carry all sorts of database co=
nnection and Statement state with them, so exposing that through a callback=
is a layer violation and a flat-out request for disaster. Don't do it.
... which is why I need the unmodifiable "view" that gives the callback
read-only access to the current row. (I probably should have stated it
that way in my original post.)
Also, the _raison d'etre_ for ResultSet is to perform 'next()' and othe=
r such calls. Asking for a ResultSet on which you don't wish to do the fun=
damental operations is a red flag that you're heading the wrong way down Fu=
bar Path.
The method that's iterating through the ResultSet obviously needs to
call next(), etc. As I said in my original post, the callback method
"has no business" changing the state of the ResultSet.
The standard approach is to unwrap the data from the ResultSet into an =
object within your domain model and pass that back. The object with the ca=
llback almost certainly doesn't want the relational model but a domain mode=
l anyway.
In fact, the class with the callback is a subclass of the class with
the "iterator" method, and it will "want" the relational model if I
write it that way.
The "iterator" method is also very general, so there isn't any way to
construct a particularly useful object for it to pass back. I could use
a Map<String,Object> or an Object[], but that's about it.
You should consider JPA. And perhaps take a course in object-oriented =
design.
Perhaps, but I'll venture that the latter is quite a conclusion to
reach from the information provided.
Are you trying to spin the paucity of information you provided as a virtue?=
Wow.
The information you presented laid out a picture of accessing a ResultSet f=
rom at least one layer away from the code that needed to see the ResultSet =
as such.
Now you blame me for reading it that way.
If you want answers that accommodate all your information, you might wish t=
o consider presenting all your information. Better yet, how about
http://sscce.org/
?
Hm?
The fact remains that there is no standard way of presenting a ResultSet as=
immutable except to extract the information from it into another structure=
, much as immutable arrays are simulated via 'return someArray.clone();'.=
Now you come back and say that you have a non-standard need for which you h=
ope to find a standard solution. As I told you, the standard solution is t=
o make a copy, transforming structures as needed.
I wish you the best of luck.
--
Lew