You can access the class index cursor using:
'make-class-cursor' for all instances and
'make-inverted-cursor' on a specific slot index
You can also use the convenience macros:
'with-inverted-cursor' and 'with-class-cursor'
These will be documented in the new manual, but for now you can look
at them
in src/elephant/classindex.lisp.
You can rely on get-instances-by-range returning ascending orders,
but you shouldn't
rely on the ordering of duplicates.
You do make the good point that I should add an option to the map
operators to return
objects in reverse order. I'll think about this.
Ian
On Apr 22, 2007, at 5:34 AM, Mac Chan wrote:
Hi Ian,
On 4/21/07, Ian Eslick <[EMAIL PROTECTED]> wrote:
This example could be done even more cleanly using the class indexing
facilities. If you feel inclined to experiment with them, this would
be a good contribution to improve the example.
Actually it turned out to be an interesting exercise. Please see the
attached file.
I have some questions, though.
* Displaying all the blog entries in descending order
Using the b-tree index, one could iterate through the entries with a
cursor as shown in the original example code:
(with-btree-cursor (curs *my-blog*)
(loop
for (m k v) = (multiple-value-list (cursor-first curs))
then (multiple-value-list (cursor-next curs))
while m
collect v))
Now after I replace the explicit index with a class index, I couldn't
figure out how to use cursor on the class index. Maybe there's an
undocumented way to do this?
(loop
for v in
(sort (get-instances-by-class 'blog-entry) #'> :key #'entry-date)
collect v)
Now I have to get all the instances from that class. I assumed the
result returned by get-instances-by-class is just a list of OID's?
Elephant will only fetch the values when I use the accessor functions
on the OID handle, right?
Another downside with this approach is that I need to manually sort
the entries (I guess if my assumption from the previous paragraph is
true, then I'm really paying the penalty here if there are a lot of
entries to sort) whereas in the original code it's taken care of by
the cursor API.
* Displaying all the blog entries for a particular month in descending
order
(with-btree-cursor (curs (get-index *my-blog* 'month))
(loop
for (m k v) = (multiple-value-list
(cursor-set curs month-year))
then (multiple-value-list (cursor-next-dup curs))
while m
collect v))
(loop
for v in
(multiple-value-bind (start end)
(the-month-range month-year)
(nreverse (get-instances-by-range 'blog-entry 'date start end)))
collect v)
In this case both the cursor and the get-instances-by-range API are
rather easy to use. Unlike get-instances-by-class,
get-instances-by-range seems to always return the result sorted, is
that correct?
When I first read the documenation, I couldn't think of a use case for
cursor-next-dup. It is really handy as shown in the original example.
* Displaying href links to previous blog entries (month/year)
(loop
for (m k v) = (multiple-value-list (cursor-first curs))
then (multiple-value-list (cursor-next-nodup curs))
while m
collect v)))
(let ((month-year))
(loop
for entry in (get-instances-by-class 'blog-entry) do
(multiple-value-bind (second minute hour date month year)
(decode-universal-time (entry-date entry))
(declare (ignore second minute hour date))
(pushnew (encode-universal-time 0 0 0 1 month year) month-year)))
(sort month-year #'>))
Now the cursor API is much easier to use in this case. Again, since I
can't figure out a way to use cursor on the class index, I need to go
through _all_ the entries, collect and sort the date myself.
I believe I'm not using this stuff right, so please show us the
way :-)
Regards,
-- Mac
<hunchentoot.blog.lisp>
_______________________________________________
elephant-devel site list
elephant-devel@common-lisp.net
http://common-lisp.net/mailman/listinfo/elephant-devel
_______________________________________________
elephant-devel site list
elephant-devel@common-lisp.net
http://common-lisp.net/mailman/listinfo/elephant-devel
