struct-maps and key removal

[Richard Newman]

Something I keep bumping into: the SQL library returns rows as struct-
maps. Often I want to do things like rename keys (:foo_bar => :foo-
bar), strip out :id columns, etc.

Nope!

java.lang.Exception: Can't remove struct key

Any opinions on returning a hash-map after 'removing' (remember, these
are persistent data structures) a key from a struct-map? They print
the same, act the same... except in this instance.

[ataggart]

So have (dissoc mystruct :akey) act like (dissoc (into {}
mystruct) :akey)? Sounds reasonable to me.

[Garth Sheldon-Coulson]

I wonder if the fact that this currently doesn't work the way you want it to is a necessary consequence of structural sharing and the desired performance guarantees of dissoc.

In other words: I wonder if, on account of how struct maps are implemented (sort of like a vector, right?), the only way to get a hash map with one key gone is to do a linear copy of the struct map.

If that's true (I have no idea if it is!), then there might be a reason for forcing you to do

(dissoc (into {} mystruct) :akey)

explicitly. Namely, it preserves the performance guarantees of dissoc.

Garth

--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clo...@googlegroups.com
Note that posts from new members are moderated - please be patient with your first post.
To unsubscribe from this group, send email to
clojure+u...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en

[Sean Devlin]

Isn't the whole point of a struct that it guarantees that certain keys
are present? I think the current behavior is correct.

On Dec 15, 5:21 am, Garth Sheldon-Coulson <g...@mit.edu> wrote:
> I wonder if the fact that this currently doesn't work the way you want it to
> is a necessary consequence of structural sharing and the desired performance
> guarantees of dissoc.
>
> In other words: I wonder if, on account of how struct maps are implemented
> (sort of like a vector, right?), the only way to get a hash map with one key
> gone is to do a linear copy of the struct map.
>
> If that's true (I have no idea if it is!), then there might be a reason for
> forcing you to do
>
> (dissoc (into {} mystruct) :akey)
>
> explicitly. Namely, it preserves the performance guarantees of dissoc.
>
> Garth
>

> On Tue, Dec 15, 2009 at 5:05 AM, ataggart <alex.tagg...@gmail.com> wrote:
>
> > On Dec 14, 11:23 pm, Richard Newman <holyg...@gmail.com> wrote:
> > > Something I keep bumping into: the SQL library returns rows as struct-
> > > maps. Often I want to do things like rename keys (:foo_bar => :foo-
> > > bar), strip out :id columns, etc.
>
> > > Nope!
>
> > >    java.lang.Exception: Can't remove struct key
>
> > > Any opinions on returning a hash-map after 'removing' (remember, these
> > > are persistent data structures) a key from a struct-map? They print
> > > the same, act the same... except in this instance.
>
> > So have (dissoc mystruct :akey) act like (dissoc (into {}
> > mystruct) :akey)?  Sounds reasonable to me.
>
> > --
> > You received this message because you are subscribed to the Google
> > Groups "Clojure" group.
> > To post to this group, send email to clo...@googlegroups.com
> > Note that posts from new members are moderated - please be patient with
> > your first post.
> > To unsubscribe from this group, send email to

> > clojure+u...@googlegroups.com<clojure%2Bunsu...@googlegroups.com>

[Richard Newman]

> Isn't the whole point of a struct that it guarantees that certain keys
> are present?

Oh, I agree -- dissoc on a struct-map would return a non-struct-map.
I'm not suggesting that struct-maps should be able to be dissociated
from one of their keys, I'm wondering whether it's better to return an
array-map or hash-map from dissoc rather than blow up.

(Or whether there should be a `without-keys` function that behaves
much like dissoc but works on any map, and have rename-keys and other
"implicit dissoc" functions use that instead. rename-keys on a struct-
map's core keys implies an abandonment of struct-ness, no?)

I suppose the question is "are struct-maps 'structs' or
maps?" (speaking ontologically). If they're structs, then it's happy
coincidence that the map get/assoc functions work on them, and
slightly confusing that they print as ordinary maps. If they're maps,
where their struct-ness is an optimization, then dissoc should work in
the same way that assoc on an array-map can sometimes return a hash-map.

The main reason I bring this up is that I had no reason to suspect
that the maps returned from my SQL queries were actually struct-maps.
They printed like any other map, and work correctly for every non-
removing map operation. Having some subset of maps error on core API
operations seems opposed to abstraction-oriented programming. It trips
my disgust switch to see code like

(rename-keys
(if (struct-map? m)
(into {} m)
m)
{:foo :bar})

[Rich Hickey]

Yes. The behavior for the new deftypes, when used as maps, is as you
desire - a dissoc of a field key yields an ordinary map (no longer of
the deftype type).

Enhancements to structs are on hold pending analysis of their utility
in light of deftypes.

Rich