inst/script/test_PkgFreeExamples.R

In Bioconductor/OrganismDbi: Software to enable the smooth interfacing of different database packages

## I want to make it so that the internal object stores the path.
## It can still store the name, but the path is the important thing that we need to stash internally.
## And the constructor can still use named object (and I think it should), but then internally when it's setting up the object, that needs to call dbfile(object) to get the path that it needs (and error informatively if that can't be done).

## The 1st decision I still have to make: keep the names or not?  I think keep them (for now), since I will have them at construction time anyways and this allows me to keep the internals generic and still have them make some human sense when testing (instead of things like object1 or 'OrgObject').

## I will need a helper function for the makeXXX code to allow me to call dbfile() on all the nodes. - this may already exist for select etc.

## I need to store the filepaths for these things and the data.frame that I currently am using is not really up to the task...  So probably I should make a list object with the names and the filepaths in a second spot?  But (importantly), I don't need to ask the user for this information.  This is just what will be stored (as an object).  So instead of a data.frame, I will now store a list that contains 1] a data.frame and 2] a vector with the file paths...




## TO Convert over:
## 1- To start I will just make a modded (example object in a test package).
## load('graphData.rda')
## resources <- c(GO.db=dbfile(GO.db), org.Hs.eg.db=dbfile(org.Hs.eg.db), TxDb.Hsapiens.UCSC.hg19.knownGene=dbfile(TxDb.Hsapiens.UCSC.hg19.knownGene))
## graphInfo <- list(graphData=graphData, resources=resources)
## save(graphInfo, file='graphInfo.Rda')
## -done

## 2- Then modify the constructor to expect and load from this kind of
## saved list object. - done


## 3- update .loadOrganismDbiPkg?  And ALSO make a .loadOrganismDb()
## function to create the object from a saved piece of data. - done
## BUT not yet tested...

## 4- Then change makeOrganismPackage to look up and save this data when users make a package.
## Done but not tested.



## Here is some test code:
library(OrganismDbi)

gd <- list(join1 = c(GO.db="GOID", org.Hs.eg.db="GO"),
           join2 = c(org.Hs.eg.db="ENTREZID",
             TxDb.Hsapiens.UCSC.hg19.knownGene="GENEID"))

## debug(OrganismDbi:::.extractDbFiles)
## debug(OrganismDbi:::makeOrganismPackage)

makeOrganismPackage(pkgname = "Homo.saps",
                    graphData = gd,
                    organism = "Homo saps",
                    version = "1.0.0",
                    maintainer = "Bioconductor Package Maintainer <maintainer@bioconductor.org>",
                    author = "Bioconductor Core Team",
                    destDir = ".",
                    license = "Artistic-2.0")























################################################################################
################################################################################
################################################################################
################################################################################


## Old example
library(OrganismDbi)


## debug(OrganismDbi:::.mungeGraphData)
## Bug fixed. - But ANOTHER one still lurks!

gd <- list(join1 = c(GO.db="GOID", org.Hs.eg.db="GO"),
           join2 = c(org.Hs.eg.db="ENTREZID",
             TxDb.Hsapiens.UCSC.hg19.knownGene="GENEID"))

makeOrganismPackage(pkgname = "Homsaps",
  graphData = gd,
  organism = "Homo sapiens",
  version = "1.0.0",
  maintainer = "Pkg Maintainer<maintainer@somewhere.org>",
  author = "Some Body",
  destDir = ".",
  license = "Artistic-2.0")


## ## This remaining bug causes this to fail: :(
## ## first an example that *will* work (traditional)
## gd <- list(join1 = c(GO.db="GOID", org.Hs.eg.db="GO"))
## makeOrganismPackage(pkgname = "sapiens",
##                     graphData = gd,
##                     organism = "Homo sapiens",
##                     version = "1.0.0",
##                     maintainer = "Pkg Maintainer<maintainer@somewhere.org>",
##                     author = "Some Body",
##                     destDir = ".",
##                     license = "Artistic-2.0")



## This is not something that I want to support? IOW, I think I want to be able to make OrganismDb objects, but NOT to make packages (unless there are real package depenedencies)
## and then one that requires we not depend on packages...
## debug(OrganismDbi:::.initPkg)
## debug(OrganismDbi:::.loadOrganismDbiPkg)
library(OrganismDbi)
library(AnnotationHub)
ah <- AnnotationHub()
puffer <- ah[['AH13961']]
##gd <- list(join1 = c(GO.db="GOID", puffer="GO"))

gd <- list(join1 = c(GO.db="GOID", puffer="GO"),
           join2 = c(puffer="ENTREZID",
             TxDb.Hsapiens.UCSC.hg19.knownGene="GENEID")) ## irrelevant TxDb
makeOrganismPackage(pkgname = "puff",
                    graphData = gd,
                    organism = "Tetraodon nigroviridis",
                    version = "1.0.0",
                    maintainer = "Pkg Maintainer<maintainer@somewhere.org>",
                    author = "Some Body",
                    destDir = ".",
                    license = "Artistic-2.0")

## then install gives me an error because its trying to install it without access to puffer....
install.packages("./puff/", repos=NULL)



####################################################################
## Lets make a simpler test that just uses the constructor:
## We need to be able to make and load this without an entire package...

## Here is what shows up in zzz.R (from .onLoad hook):
## So its just saving a data.frame (nothing special - but I *DO* need it...
load(system.file("data","graphData.rda",package='puff'))
## OrganismDbi:::.loadOrganismDbiPkg(pkgname='puff',
##                                  graphData=graphData)

## So basically: export this constructor here so that it can be called separately.
## Note: dbType argument is not even used...
## debug(OrganismDbi:::OrganismDb)
debug(OrganismDbi:::.initPkg)
debug(OrganismDbi:::.addLocalPkgsToNamespace)
debug(OrganismDbi:::.loadOrganismDbiPkg)

## install.packages("./puff/", repos=NULL)
library("puff")
puff

## puff <- OrganismDbi:::OrganismDb(graphData=graphData)
columns(puff)







########################################################################
## Now lets make an example with an Annotationhub resource, but that does not store the resource locally... (instead looks it up each time from the hub cache).

## But before I can do that, I need to make another change...  Basically, I need to change the internals of the object so that instead of storing the 'names' of the objects (which could change or eventually clash etc.) I instead use dbfile() to get (and store) the locations of the database files that exist for each one.  For those cases where an object does not yet have a file location, THEN we will do what we are doing now (and save it to the package/rely on it being in the path).

## so the plan will be to rely more on something like this:
## Does it have a value for dbfile()? (this covers installed packages AND AnnotHub objects)

## I am not sure if I can support objects that are only on the path (and unsaved):
## Ideally if it is *only* on the path (IOW there is no value for dbfile)? If so, then save it to the package (only can offer this saving option in case where they call makeOrganismPackageFromXXX() family... So users will be able to make an object that can't load the next time (if saved)

## So that would mean that the heuristic would be:
## 1 - is it on the path already? (exists())
## 2 - if not call loadDb on the file.path in the object

## And internally, the object would store either a name OR a character vector (IOW file path)

## Recommendation in that case will be to them to use the constructor only with reliable resources (packages or AnnotationHub objects).


## Alternatively I could just REQUIRE that all objects all be based on objects that have been saveDb/loadDb cycled...  But I think I want to keep the current flexibility.




## WHAT I REALLY WANT:

## all resources in a OrganismDb object must point to saved DBs (file paths only).
## if the end user calls makeOrganismPackageFromXXX() and passes in an object that is not saved to disc, it will be saved to their package
## if the user tries to make an object with something that is not saved to disc (using the constructor) then they should get a warning (but the constructor should basically just call it with exists() etc. (as it basically does now). 


## ??? Also the arguments (even for the general case) need to be a little different since I will now need to pass in actual objects (and not just pairs of strings).  This means I need to modify this to also have the actual objects passed in? Or maybe that's not all that important (since internally they will be object whether the actual object or just a name is passed in)...

## Here is what these are currently:
## gd <- list(join1 = c(GO.db="GOID", puffer="GO"),
##            join2 = c(puffer="ENTREZID",
##              TxDb.Hsapiens.UCSC.hg19.knownGene="GENEID")) ## irrelevant TxDb