Thanks!

I updated the documentation for alternation and for lazy search

OK, changed and pushed to master. The value of 1000 in lazy was also used in two
places, so I've replaced it with a symbolic constant.

Can someone else update the documentation?

No objections. :-)

Gabi, any objections?

I've updated the documentation, but I agree with Malte about changing the default 
value of boost in alternation open list.

Some comments on the docs:

 * The docs need not repeat the default if it's already given in the signature
   -- it's enough to maintain this in one place, otherwise this info tends to
   not get updated consistently.

 * There's a bug in one of the examples: the version where boost is not
   specified ("search h1=eval1 --search eager_greedy(h1, preferred=(h1))")
   is *not* equivalent to the following call that uses alt(), since eager_greedy
   has a default of 0 for boost, but alt has a default of 1000. Hence, in
   cases where no value is given, the value of 0 for alt would need to be
   given explicitly.

   That's a somewhat subtle and surprising issue, and I suggest to address that
   by changing the "boost" default of alt to 0 (while keeping the 1000 default
   of lazy_greedy).

Committed this with two changes:
 * There was a comment adding that looked to be interrupted in the middle,
   so I completed it.
 * The default value for the eager boost was 1000, but it should be 0 as
   discussed below. Changed it to 0.

This is now in the master. Erez, can you add documentation for this for
eager_greedy and mark as resolved?

This is now in issue134 in my repository

This is not yet merged, right? I just checked your repository, and this is
tangled with all the problem adapter changes so that I cannot easily merge this.

Since this is a small change, it'd be good if this weren't held up until the
problem adapter is merged. Hence, the way forward would be for you to create a
new branch "issue134" off the current default branch and make the change there.

Alternatively, we can also wait with this until the problem adapter is
integrated, but that can take a while.

Sure, I'll update this issue then and you'll get a notification. Could be a
while though; teaching started last week, things are a bit busy.

Done.
Malte - let me know when you merge.

If I understand these terms correctly, I'd go with "boost_preferred" always and
get rid of boost_last_used() -- this is what is described and evaluated in the
literature.

Extensions may be fine at a later point, but they should have a proven benefit.
Too many options is bad for maintenance.

I changed the default value of boost to 0 in my repository.
Let me know when you merge this, and I'll update the wiki.
Also - I used boost_last_used(), while in lazy search we use boost_preferred().
Maybe we should make this a parameter?

> I just noticed that boosting is not used at all with eager search.
> This is an easy fix, I'm just wondering if there's a reason it's not there.

Ask yourself, I think you implemented eager search. :-) (I originally only
implemented A*, for which there were neither boosting nor preferred operators
because we wanted it to be optimal.)

Boosting should be offered as an option for eager search, but according to the
ICAPS 2009 paper by Silvia and me, it should be off by default for eager search.

The fix is available on my hg repository in Issue134 branch

I just noticed that boosting is not used at all with eager search.
This is an easy fix, I'm just wondering if there's a reason it's not there.