Parsing a Simple Chemical Formula

Hi David & others...

I did find the function you recommended, plus, it's even easier (but a 
little hidden in the doc): >element(form, "mass").  But, this uses the 
atomic masses from the periodic table, which are weighted averages of 
the isotopes of each element.  What I'm doing actually involves mass 
spectrometry, so I need the isotope masses, which are integers (think 
12C, 13C, 14C, but the periodic table says 12.011 reflecting the 
relative abundances).  I used Gabor's solution and got my little 
function humming.  Plus, I have several things to read through from 
the various recommendations.

Thanks again, Bryan

On Dec 26, 2010, at 10:21 PM, David Winsemius wrote:

On Dec 26, 2010, at 8:28 PM, Bryan Hanson wrote:

Thanks Spencer, I'll definitely have a look at this package and it's 
vignettes.  I believe I have looked at it before, but didn't catch 
it on this particular search.  Bryan

Using the thermo list that the makeup function accesses to get its 
valid atomic symbols one can arrive at the the answer you posited 
would be too difficult in you first posting, the atomic weight from 
the formulae:

str(thermo$element)

'data.frame':    130 obs. of  6 variables:
$ element: chr  "Z" "O" "H" "He" ...
$ state  : chr  "aq" "gas" "gas" "gas" ...
$ source : chr  "CWM89" "CWM89" "CWM89" "CWM89" ...
$ mass   : num  0 16 1.01 4 20.18 ...
$ s      : num  -15.6 49 31.2 30.2 35 ...
$ n      : int  1 2 2 1 1 1 1 1 2 2 ...

patts <- paste("^", rownames(makeup(form)), "$", sep="")
makuform<- makeup(form)
makuform$amass <- sapply(patts, function(x) {return( thermo$element[ 
grep(x, thermo$element[[1]])[1], "mass"])}  )
sum(makuform$amass *makuform$count)
# [1] 167.0457

On Dec 26, 2010, at 8:16 PM, Spencer Graves wrote:

p.s.  help(pac=CHNOSZ) reveals that this package has 3 vignettes.  
I have not looked at these vignettes, but most vignettes provide 
excellent introductions (though rarely with complete coverage) of 
important capabilities of the package.  (The 'sos' package includes 
a vignette, which exposes more capabilities than the example below.)


######################
   Have you considered the 'CHNOSZ' package?

makeup("C5H11BrO" )

count
C      5
H     11
Br     1
O      1


   I found this using the 'sos' package as follows:


library(sos)
cf <- ???'chemical formula'
found 21 matches;  retrieving 2 pages
cf


   The print method for "cf" opened the results in a web browser, 
which showed that the "CHNOSZ" package had 14 of these 11 matches, 
and the other 7 were in 7 different packages.  Moreover, the 
"CHNOSZ" package is devoted to "Chemical Thermodynamics and 
Activity Diagrams" and provides many more capabilities that might 
interest you.


   Hope this helps.
   Spencer


On 12/26/2010 5:01 PM, Bryan Hanson wrote:

Well let me just say thanks and WOW!  Four great ideas, each 
worthy of
study and I'll learn several things from each.  Interestingly, these
solutions seem more general and more compact than the solutions I
found on the 'net using python and perl.  More evidence for the power
of R!  A big thanks to each of you!  Bryan

On Dec 26, 2010, at 7:26 PM, Gabor Grothendieck wrote:

On Sun, Dec 26, 2010 at 6:29 PM, Bryan Hanson <hanson at depauw.edu> 
wrote:

Hello R Folks...

I've been looking around the 'net and I see many complex 
solutions in
various languages to this question, but I have a pretty simple need
(and I'm
not much good at regex).  I want to use a chemical formula as a
function
argument.  The formula would be in "Hill order" which is to list C,
then H,
then all other elements in alphabetical order.  My example will 
have
only a
limited number of elements, few enough that one can search directly
for each
element.  So some examples would be C5H12, or C5H12O or C5H11BrO
(note that
for oxygen and bromine, O or Br, there is no following number
meaning a 1 is
implied).

Let's say

form <- "C5H11BrO"

I'd like to get the count of each element, so in this case I 
need to
extract
C and 5, H and 11, Br and 1, O and 1 (I want to calculate the 
molecular
weight by mulitplying).  Sounds pretty simple, but my experiments
with grep
and strsplit don't immediately clue me into an obvious 
solution.  As
I said,
I don't need a general solution to the problem of calculating 
molecular
weight from an arbitrary formula, that seems quite challenging, 
just
a way
to convert "form" into a list or data frame which I can then do the
math on.

Here's hoping this is a simple issue for more experienced R users!
TIA,

This can be done by strapply in gsubfn.  It matches the regular
expression to the target string passing the back references (the
parenthesized portions of the regular expression) through a 
specified
function as successive arguments.

Thus the first arg is form, your input string.  The second arg is 
the
regular expression which matches an upper case letter optionally
followed by lower case letters and all that is optionally 
followed by
digits.  The third arg is a function shown in a formula
representation. strapply passes the back references (i.e. the 
portions
within parentheses) to the function as the two arguments.  Finally
simplify is another function in formula notation which turns the
result into a matrix and then a data frame.  Finally we make the
second column of the data frame numeric.

library(gsubfn)

DF <- strapply(form,
"([A-Z][a-z]*)(\\d*)",
~ c(..1, if (nchar(..2)) ..2 else 1),
simplify = ~ as.data.frame(t(matrix(..1, 2)), stringsAsFactors =
FALSE))
DF[[2]] <- as.numeric(DF[[2]])

DF looks like this:

DF

V1 V2
1  C  5
2  H 11
3 Br  1
4  O  1



-- 
Statistics & Software Consulting
GKX Group, GKX Associates Inc.
tel: 1-877-GKX-GROUP
email: ggrothendieck at gmail.com

David Winsemius, MD
West Hartford, CT