Add $,pSet method to easily access columns in the phenoData

[jorainer]

• Add $, $&lt;- and pData<- methods to easily get and replace/add columns to the
  phenodata.
• Add related unit tests and add documentation.

Basically, It facilitates accessing pheno data columns by just calling pset$pheno_col. I find that pretty convenient e.g. in ExpressionSet objects and alike, so I thought it might be nice to have it in MSnbase as well.
Also I added replace methods for the pData.

[codecov-io]

Codecov Report

Merging #203 into master will increase coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #203      +/-   ##
==========================================
+ Coverage   71.72%   71.74%   +0.01%     
==========================================
  Files          67       67              
  Lines        7165     7170       +5     
==========================================
+ Hits         5139     5144       +5     
  Misses       2026     2026

Impacted Files	Coverage Δ
R/methods-pSet.R	81.19% <100%> (+0.44%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d67ab4d...77a8bd6. Read the comment docs.

[lgatto]

Thanks @jotsetung, I think this will prove very useful. I'll wait for the new release to merge.

On a similar note, I also thought it would be useful to be able to change the behaviour of $ through a MSnbase option, where it could either link to pData (default) or fData. The latter is much more useful in our spatial proteomics work in pRoloc, where we so often need to access feature variables in our MSnSets.

[jorainer]

Don't you think that might become a little confusing if $ is used to access both rows or columns? What about $$ to access fData and $ for pData?

[lgatto]

Hmm, $$, that is an interesting suggestion. Or what about £ (no, that's a joke) :-). Ok, let's go for $$, as that will even bypass the need to setting an option. Whoever has time for this is welcome to have a go, but won't be merged before the new release.

[jorainer]

I checked the $$ option but that does not seem to be that straight forward. Didn't find where $ is defined to check how and if a function named $$ could be defined to have the same behavior (i.e. that it is possible to call it like x$$y instead of $$(x, y).
@lgatto do you know by chance how and where $ is defined?

[sgibb]

I am not sure that it is possible to use $$ as function. We can define other operators:

library("MSnbase")
data(msnset)

setGeneric("%fd%", function(object, name)standardGeneric("%fd%"))
setGeneric("%pd%", function(object, name)standardGeneric("%pd%"))

setMethod("%fd%", "MSnSet", function(object, name)fData(object)[, as.character(substitute(name))])
setMethod("%pd%", "MSnSet", function(object, name)pData(object)[, as.character(substitute(name))])

msnset %fd% ProteinAccession
#  [1] BSA     ECA1422 ECA4030 ECA3882 ECA1364 ECA0871
#  [7] ECA4512 ECA4513 ECA3969 ECA3082 ECA1032 ECA1294
# [13] ECA4514 ECA1104 ECA3356 ECA4037 ECA0621 ECA1093
# [19] ECA0452 ENO     ECA2391 ECA0435 ECA1362 ECA1363
# [25] ECA3566 ECA0435 ECA0691 ECA3566 ECA3377 ECA0978
# [31] ECA2831 ECA4514 ECA0469 ECA4514 ECA4514 ECA0172
# [37] ECA0631 ECA3349 ECA0469 ECA4514 ENO     ENO
# [43] ECA4026 ECA3929 ECA4514 ECA4026 ECA4013 BSA
# [49] BSA     ECA2186 ENO     ECA3349 ECA2421 ECA1443
# [55] ECA3175
# 40 Levels: BSA ECA0172 ECA0435 ECA0452 ... ENO

msnset %pd% mz
# [1] 114.1 115.1 116.1 117.1

(%fd<-%, %pd<-% would be possible, too)
Of course tab completion is not working. Maybe we can hack something via ?rcompgen. But I am not sure that is worth the effort: msnset %fd% ProteinAccession is IMHO no real improvement over fData(msnset)$ProteinAccession.

[sgibb]

`$$` <- function(x, y) { x[[y]] }
l <- list(a=1, b=2)
l$$b
# Error: unexpected '$' in "l$$"
$$(l, "a")
# Error: unexpected '$' in "$"

[jorainer]

Yes, that was what I realized too, defining $$ is tricky.

Regarding the %fd% etc operators, it would be possible, but might not be intuitive. What I mean is that many users that use e.g. ExpressionSet, SummarizedExperiment etc are familiar with the use of $ to access phenodata columns. That was also the first thing I missed in the pSet. I would not implement a completely new shortcut to do the same (i.e. the %pd%).

Eventually we just add the $ and use the old fashioned fData(msd) to access feature annotations?

[lgatto]

Have a look in Biobase to see how things have been implemented for $.

[jorainer]

That's exactly our implementation for the $ in MSnbase. The problem is that $$ does not work in a similar fashion (as @sgibb pointed out).

[lgatto]

Following up from @sgibb's example:

[1] 1
> `$$` <- function(x, y) { x[[y]] }
> l <- list(a=1, b=2)
> `$`(l, "a")
[1] 1
> `$$`(l, "a")
[1] 1

I think the problem we have is that R can't parse $$ in l$$a, while it knows how to handle $.

[lgatto]

An alternative would be to use @, but I always stumble on the wrong signature

> .Foo <- setClass("Foo", slots = c(a = "numeric"))
> foo <- .Foo(a = 1:10)
> names(foo@a) <- letters[1:10]
> `$`
.Primitive("$")
> getGeneric("$")
standardGeneric for "$" defined from package "base"
function (x, name) 
standardGeneric("$", .Primitive("$"))
<bytecode: 0x1355f40>
<environment: 0x1221b00>
Methods may be defined for arguments: x
Use  showMethods("$")  for currently available ones.
> setMethod("$", "Foo", function(x, name) x@a[name])
[1] "$"
> foo$a
a 
1 

but

> getGeneric("@")
NULL
> `@`
.Primitive("@")
> setMethod("@", "Foo", function(x, name) x@a[name])
Error in setGeneric(f, where = where) : 
  ‘@’ dispatches internally;  methods can be defined, but the generic function is implicit, and cannot be changed.
> setMethod("@", "Foo", function(object, name) object@a[name])
Error in setGeneric(f, where = where) : 
  ‘@’ dispatches internally;  methods can be defined, but the generic function is implicit, and cannot be changed.

[jorainer]

I think @ would also confuse people since it's the accessor for the slots of the class.

[lgatto]

Yes, but sometime the perfect is the enemy of good. I am unsure that $$ would ever work, hence may be better @ than nothing, ... My idea or using £ might become the least bad solution in the end!

[lgatto]

I am going to merge this. Issue #209 follows up the discussion to directly access fData columns.