rewrite topN and remove getTopIdx/subsetBy

[sgibb]

This PR replaces getTopIdx and subsetBy by .topIdx.

.topIdx runs the summarise function just once over the whole matrix (instead of groupwise as getTopIdx did) and returns global row indicies that could be directly used to subset the matrix (no need for subsetBy anymore).
The code is much easier to read and it greatly improves the runtime (in this example 4x but in our current synapter set with 30000 peptides it is nearly 40x):

library("git2r")
library("devtools")

repo <- repository("MSnbase")

checkout(repo, branch="master")
load_all("MSnbase")
data("msnset2")
system.time({mas <- topN(msnset2, groupBy = fData(msnset2)$accession, n = 3) })
#    user  system elapsed
#   0.092   0.000   0.091

checkout(repo, branch="topN")
load_all("MSnbase")
system.time({top <- topN(msnset2, groupBy = fData(msnset2)$accession, n = 3) })
#    user  system elapsed
#   0.019   0.000   0.018

all.equal(mas, top)
#  [1] "Attributes: < Component “.__classVersion__”: Component “R”: Mean relative difference: 0.6666667 >"
#  [2] "Attributes: < Component “.__classVersion__”: Component “Biobase”: Mean relative difference: 0.2352941 >"
#  [3] "Attributes: < Component “.__classVersion__”: Component “pSet”: Mean relative difference: 1 >"
#  [4] "Attributes: < Component “processingData”: Attributes: < Component “processing”: Lengths (2, 3) differ (string compare on first 2) > >"
#  [5] "Attributes: < Component “processingData”: Attributes: < Component “processing”: 1 string mismatch > >"
#  [6] "Attributes: < Component “qual”: Attributes: < Names: 3 string mismatches > >"
#  [7] "Attributes: < Component “qual”: Attributes: < Length mismatch: comparison on first 3 components > >"
#  [8] "Attributes: < Component “qual”: Attributes: < Component 2: Attributes: < Modes: list, NULL > > >"
#  [9] "Attributes: < Component “qual”: Attributes: < Component 2: Attributes: < names for target but not for current > > >"
# [10] "Attributes: < Component “qual”: Attributes: < Component 2: Attributes: < current is not list-like > > >"
# [11] "Attributes: < Component “qual”: Attributes: < Component 2: Lengths (1, 0) differ (string compare on first 0) > >"
# [12] "Attributes: < Component “qual”: Attributes: < Component 3: Modes: character, numeric > >"
# [13] "Attributes: < Component “qual”: Attributes: < Component 3: target is character, current is numeric > >"

## synapter
library("synapter")
r <- readRDS("kuharev2015/synapter2/output/combMSnSet.RDS")
g <- fData(r)$protein.Accession.S130423_05

repo <- repository("MSnbase")

checkout(repo, branch="master")
load_all("MSnbase")
system.time({mas <- topN(r, groupBy = g, n = 3) })
#       User      System verstrichen
#     37.928       0.364      38.297

checkout(repo, branch="topN")
load_all("MSnbase")
system.time({top <- topN(r, groupBy = g, n = 3) })
#    user  system elapsed
#   0.088   0.000   0.089

Obviously the MSnSet are not identical anymore because I use subset instead of recreation now (but recreation would also be possible if wished).

There is also a new function .summariseRow that just run apply(x, 1, fun, ...) in most cases. But if the user want to use sum/mean it uses rowSums/rowMeans. We could use .summariseRow in other places too (if you like). Please note that summariseRow could easily be fouled: summariseRow(x, fun=function(y)sum(y)) will use apply instead of rowSums.

[codecov-io]

Codecov Report

Merging #199 into master will decrease coverage by 0.05%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #199      +/-   ##
==========================================
- Coverage    71.8%   71.75%   -0.06%     
==========================================
  Files          67       67              
  Lines        7182     7165      -17     
==========================================
- Hits         5157     5141      -16     
+ Misses       2025     2024       -1

Impacted Files	Coverage Δ
R/utils.R	61.55% <100%> (+0.42%)	⬆️
R/methods-MSnSet.R	56.22% <100%> (-1.92%)	⬇️

---

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 cb2fb86...bfddd5f. Read the comment docs.

[lgatto]

@sgibb thank you very much! Could you document this in the news file, please, possibly with the PR number. That will make it easier if, in the future, we need to track that change back.