Enable velocyto.R support both hdf5r and h5

[Puriney]

Motivation
The installation of velocyto.R cannot work mainly due to one of the dependencies, h5; this was also reported in these issues #42, #30, and #13 (#13 (comment)).

Here Yun enables the velocyto.R to support both hdf5r and h5, keeps the results consistent and avoids messing up the original code structure. In addition, this PR leaves flexibility for the author to decide if to keep h5 or exclusively to use hdf5r.

What's Changed

1. As suggested by the author (Error installing velocyto.R #30 (comment)), the two affected functions read.loom.matrices() and read.gene.mapping.info() were fixed so that now they support both the hdf5r and h5. In details, they are both added one parameter, named engine, to select either 'hdf5r' or 'h5'. Default value is 'hdf5r'.
2. In the DESCRIPTION file, 'h5' was changed to 'Suggests' section and 'hdf5r' was changed to 'Imports' section, because the h5 package is officially deprecated in favor of the hdf5r package. (See notice here).
3. The 'man/*.Rd' manuals were automatically generated and so updated by roxygen2. The contents were consistent except that one error existed in 'man/find.ip.sites.Rd' was fixed. roxygen2 picked up a deprecated function (i.e., t.generate.ip.mask) and replaced it with the correct one (i.e., find.ip.sites).

Repeatable Demo
Use the official tutorial data to run the repeatable demo. The results were same and consistent.

1. read.loom.matrices()

# 10X43_1.loom from: http://pklab.med.harvard.edu/velocyto/DG1/10X43_1.loom
fpath_loom <- 'data/10X43_1.loom'
ldat_h5 <- read.loom.matrices(fpath_loom, engine = 'h5')
ldat_hdf5r <- read.loom.matrices(fpath_loom, engine = 'hdf5r')
all.equal(ldat_h5, ldat_hdf5r)
# TRUE

2. read.gene.mapping.info()

library(parallel)
# http://pklab.med.harvard.edu/velocyto/notebooks/R/chromaffin2.nb.html
ip.mm10 <- readRDS(url("http://pklab.med.harvard.edu/velocyto/chromaffin/ip.mm10.rds"))
gene.info <- read.gene.mapping.info(
  "data/velocyto/dump/onefilepercell_A1_unique_and_others_J2CH1.hdf5",
  internal.priming.info=ip.mm10,
  min.exon.count=10,
  engine='h5')
gene.info2 <- read.gene.mapping.info(
  "data/velocyto/dump/onefilepercell_A1_unique_and_others_J2CH1.hdf5",
  internal.priming.info=ip.mm10,
  min.exon.count=10,
  engine='hdf5r')

all.equal(gene.info, gene.info2)
# [1] "Component “info”: Component “is_intron”: 'current' is not a factor"    
# [2] "Component “info”: Component “is_last3prime”: 'current' is not a factor"
# [3] "Component “info”: Component “strandplus”: 'current' is not a factor" 

all.equal(gene.info$gene.df, gene.info2$gene.df)
# TRUE

for (x in names(gene.info$info)){
  cat(x, ', ')
  cat(all.equal(gene.info[[x]], gene.info2[[x]]))
  cat('\n')
}
# chrm , TRUE
# exino , TRUE
# features_gene , TRUE
# is_intron , TRUE
# is_last3prime , TRUE
# start_end , TRUE
# strandplus , TRUE
# tr_id , TRUE
# cluster.feature.counts , TRUE

Tested Env

• Both the h5 and hdf5r can be installed
  • R version 3.4.2
  • x86_64-apple-darwin15.6.0 (64-bit)
• Version if this PR were merged is successfully installed
  • R version 3.5.1
  • x86_64-apple-darwin15.6.0 (64-bit)

---

Thanks @alexbarrera for following up my issue.

[Puriney]

roxygen2 picks up a deprecated function (i.e., t.generate.ip.mask) and replace it with the correct one (i.e., find.ip.sites).

[mgood2]

I get this error msg.

Error: package or namespace load failed for ??hdf5r?? in dyn.load(file, DLLpath = DLLpath, ...):
 unable to load shared object '/home3/elee4472/R/x86_64-pc-linux-gnu-library/3.5/hdf5r/libs/hdf5r.so':
  libhdf5_hl.so.100: cannot open shared object file: No such file or directory
Error: loading failed
Execution halted
ERROR: loading failed

[Puriney]

I get this error msg.

Error: package or namespace load failed for ??hdf5r?? in dyn.load(file, DLLpath = DLLpath, ...):
 unable to load shared object '/home3/elee4472/R/x86_64-pc-linux-gnu-library/3.5/hdf5r/libs/hdf5r.so':
  libhdf5_hl.so.100: cannot open shared object file: No such file or directory
Error: loading failed
Execution halted
ERROR: loading failed

The error is all about the installation of hdf5r itself. The exact same issue/problem was reported here hhoeflin/hdf5r#106. Once you have hdf5r installed, you are probably ready to install and use velocyto.R.

[pkharchenko]

Got the following warning messages in the first test:

> ldat_hdf5r <- read.loom.matrices(fpath_loom, engine = 'hdf5r')
reading loom file via hdf5r...
[1] "Couldn't delete        144115188075855873"
[1] "Couldn't delete        144115188075855875"
[1] "Couldn't delete        144115188075855878"
[1] "Couldn't delete        144115188075855880"
[1] "Couldn't delete        144115188075855882"
> all.equal(ldat_h5, ldat_hdf5r)
[1] TRUE

The result is OK though.

[pkharchenko]

OK, merged in to dev, tested, and merged all into master. Thank you very much for the fix!