chemiclj is a library for representing and processing various kinds of chemical entities, such as elements, atoms, bonds and molecules. To see chemiclj in action, let's build a model of methane (we'll see later how to do this less tediously). But first, a word about clojure programming style. Coming from a background in Java, C++, Common Lisp, etc..., one might expect that to build a model of methane, you create a molecule object, and then add atoms and bonds to it. While that's smililar to what one does in clojure (in indeed can do with clojure using the mutable data types and their operators), the clojur-ish thing to do seems to be to create new objects at each step along the way, rather than mutating the existing objects. Therefore, we create a molecule object, and, say, add a carbon to it, but instead of receiving the same object, modified in place, we receive a new object with the carbon in it. So, to build up a molecule like methane one would do:
(require '[chemiclj [core :as chem]])
;=> nil
(def methane
(let [c1 (chem/make-atom "C" "C1")
h1 (chem/make-atom "H" "H1")
h2 (chem/make-atom "H" "H2")
h3 (chem/make-atom "H" "H3")
h4 (chem/make-atom "H" "H4")]
(chem/add-bond
(chem/add-bond
(chem/add-bond
(chem/add-bond
(chem/add-atom
(chem/add-atom
(chem/add-atom
(chem/add-atom
(chem/add-atom
(chem/make-molecule)
c1)
h1)
h2)
h3)
h4)
c1 h1)
c1 h2)
c1 h3)
c1 h4)))
#'user/methane
This looks really tedious, but we'll see how to avoid this tedium later. For the moment, let's explore some simple operations on methane:
(chem/mass methane)
;=> 16.042461
(chem/exact-mass methane)
;=> 16.0313
(map name (chem/atoms methane))
;=> ("C1" "H2" "H3" "H1" "H4")
(map (comp (partial map name) chem/atoms) (chem/bonds methane))
;=> (("C1" "H4") ("C1" "H3") ("C1" "H2") ("C1" "H1"))
Ok, so it seems we do indeed have an object that represents methane and we can perform some simple operations on it. The last form is worthy of examining more closely. Notice that this is as map form, that is we'll be performing a function on each of the objects in the seq (the clojure abstraction that provides sequential access to the items in lists, vectors, maps, etc...) that we're mapping over, and accumulating the results. The seq that we'll be mapping over is (chem/bonds methane), which, as one might expect, gives us a seq of the bonds in the specified molecule. But just calling (chem/bonds methane) gives us the entire Bond object for each bond, more data than we want for the moment:
(chem/bonds methane)
;=> (#:chemiclj.bond.Bond{:_nodes [#:chemiclj.atom.Atom{:_name "C1",
:element #:chemiclj.element.Element{:atomic-number 6, :id "C", :name
"carbon", :group 14, :period 2, :mass 12.0107, :electronegativity
2.55, :max-bond-order 4}, :isotope nil, :chirality nil, :charge 0,
:hybridization nil, :aromatic nil, :explicit-hydrogen-count nil}
#:chemiclj.atom.Atom{:_name "H4", :element
#:chemiclj.element.Element{:atomic-number 1, :id "H", :name
"hydrogen", :group 1, :period 1, :mass 1.00794, :electronegativity
2.1, :max-bond-order 1}, :isotope nil, :chirality nil, :charge 0,
:hybridization nil, :aromatic nil, :explicit-hydrogen-count nil}],
:type nil, :order nil, :direction nil} ...)
I've trimmed the output for brevity's sake, but this is more than
we want at the moment. We just want the names of the atoms in the
bonds. So, now we're going to map over the bonds, extracting the
names as we go. First we need the atoms, which we can get with
(chem/atoms
(map chem/atoms (chem/bonds methane))
;=> ([#:chemiclj.atom.Atom{:_name "C1", :element
#:chemiclj.element.Element{:atomic-number 6, :id "C", :name "carbon",
:group 14, :period 2, :mass 12.0107, :electronegativity 2.55,
:max-bond-order 4}, :isotope nil, :chirality nil, :charge 0,
:hybridization nil, :aromatic nil, :explicit-hydrogen-count nil}
... }
But this is just as bad as before. We need the name of the atoms. But we can't just call name on the atoms, as what we've gotten back from the map form is a seq of seqs, rather than just a seq. We could do:
(map #(map name %) (map chem/atoms (chem/bonds methane)))
But I don't like the triple map above. It's logical that we'll have two levels of mapping, because we have a seq of seqs, but we should be able to get away with mapping a single function, which in turns maps over its items and does the right thing. Two tricks we can use are comp and partial. comp composes two functions, that is (comp f g) returns a new function that does: (fn [...] (f (g ...))). Using comp, we can map a single function that is the composition a function that maps over the atoms, giving the names of the atoms, and the function that gives us the atoms:
(map
(comp #(map name %) chem/atoms)
(chem/bonds methane))
;=> (("C1" "H4") ("C1" "H3") ("C1" "H2") ("C1" "H1"))
But now we get rid of the ugly #(map name %) form by partially applying
name to map, yielding a function that will take on argument and call (map
name
(map
(comp (partial map name) chem/atoms)
(chem/bonds methane))
(("C1" "H4") ("C1" "H3") ("C1" "H2") ("C1" "H1"))
So, there you have it. A long, complicated way to perform simple operations on one of the simplest organic molecules. In the next post we'll see how to avoid some of this tedium and start building more complex molecules, exploring more of the chemiclj API.
--