R/methods-geneCentricDbs.R

Defines functions .taxonomyId mapIds_base .cols .noSchemaCols .getDataTables .legacyCols .chooseJoinType .select testSelectArgs .testForValidKeys .testForValidCols testForValidKeytype .isSingleString .noSchemaSelect .appendGenesToTabs .legacySelect .warnAboutManyToOneRelationships .adjustForDupColNames .replaceValues resort_base .generateExtraRows .resortColumns .dropUnwantedRows .filterSuffixes .simplifyCols .extractData .generateQuery .attachDB .getDBLocs .getDBLoc .definePossibleTables .defineTables .expandCols .getChipDbFile .getOrgPkg

Documented in mapIds_base resort_base testForValidKeytype testSelectArgs

### =========================================================================
### Methods form mapping keys to table and fields, select() and
### column() methods for gene-centric Dbs
### -------------------------------------------------------------------------

## For GO, I should just define a whole distinct set.

## For Chip packages, they should be nearly the same, but we either have to do
## some extra complexity to join across DBs OR I can just do one final merge
## in R.  I may want my own .generateQuery() function for chip packages, or I
## may just want my own .extractData() function(), and to point to a different
## "x" when using chip packages.

## ALSO: I will need a way to deduce the org package that goes with each chip
## package. hgu95av2ORGPKG will (for example) get you this.  DONE:
.getOrgPkg <- function(x){
    pkgname <- packageName(x)
    orgsymbol <- sub(".db$","ORGPKG", pkgname)
    orgroot <- get(orgsymbol, getNamespace(pkgname))
    orgpkgname <- paste0(orgroot, ".db")
    get(orgpkgname, loadNamespace(orgpkgname))
}

## This gets the exact path to the chip DB.
.getChipDbFile <- function(x){
    pkgname <- packageName(x)
    pkgroot <- sub(".db$","", pkgname)
    filesymbol <- paste0(pkgroot, "_dbfile")
    get(filesymbol, loadNamespace(pkgname))()
}

## Limitation: I can only have ONE table and ONE field for each list name.
## So if we have fields like GO that should pull back multiple things, then we
## have to expand those ahead of time.
.expandCols <- function(x, cols){
  ## known expansions for cols:
  if("CHRLOC" %in% cols){ 
    after <- match("CHRLOC", cols)
    cols <- append(cols, c("CHRLOCCHR"),after) 
  }
  if("GO" %in% cols){ 
    after <- match("GO", cols)
    cols <- append(cols, c("EVIDENCE","ONTOLOGY"),after) 
  }
  if("GOALL" %in% cols){ 
    after <- match("GOALL", cols)
    cols <- append(cols, c("EVIDENCEALL","ONTOLOGYALL"),after) 
  }
  if("ORF" %in% cols && species(x)=="Saccharomyces cerevisiae"){
    after <- match("ORF", cols)
    cols <- append(cols, c("SGD"),after) 
  }  
  if("COMMON" %in% cols){ 
    after <- match("COMMON", cols)
    cols <- append(cols, c("SGD"),after) 
  } ## special case: if PROBEIDS are requested, they MUST be in front!
  if("PROBEID" %in% cols){ 
    cols <- append("PROBEID", setdiff(cols, "PROBEID"))
  }
  unique(cols)
}


## ORG PKGS CHECKLIST: 
## org.Ag.eg.sqlite           ## done
## org.At.tair.sqlite         ## done
## org.Bt.eg.sqlite           ## done
## org.Ce.eg.sqlite           ## done
## org.Cf.eg.sqlite           ## done
## org.Dm.eg.sqlite           ## done
## org.Dr.eg.sqlite           ## done
## org.EcK12.eg.sqlite        ## done
## org.EcSakai.eg.sqlite      ## done
## org.Gg.eg.sqlite           ## done
## org.Hs.eg.sqlite           ## done
## org.Mm.eg.sqlite           ## done
## org.Mmu.eg.sqlite          ## done
## org.Pf.plasmo.sqlite       ## done
## org.Pt.eg.sqlite           ## done
## org.Rn.eg.sqlite           ## done
## org.Sc.sgd.sqlite          ## done
## org.Ss.eg.sqlite           ## done
## org.Xl.eg.sqlite           ## done
## none of the above..        ## done



.defineTables <- function(x){
  class <- class(x)
  species <- species(x)
  .definePossibleTables(class, species)
}

.definePossibleTables <- function(class, species){  
  ## 1st the generic/universal things
  .defTables <- list("ENTREZID" = c("genes","gene_id"),
                     "PFAM" = c("pfam","pfam_id"),
                     "IPI" = c("pfam","ipi_id"),
                     "PROSITE" = c("prosite","prosite_id"),
                     "ACCNUM" = c("accessions","accession"),
                     "ALIAS" = c("alias","alias_symbol"),
                     "ALIAS2EG" = c("alias","alias_symbol"),
                     "ALIAS2PROBE" = c("alias","alias_symbol"),
                     "CHR" = c("chromosomes","chromosome"),
                     "CHRLOCCHR" = c("chromosome_locations","seqname"),
                     "CHRLOC" = c("chromosome_locations","start_location"),
                     "CHRLOCEND" = c("chromosome_locations","end_location"),
                     "ENZYME" = c("ec","ec_number"),
                     "MAP" = c("cytogenetic_locations","cytogenetic_location"),
                     "PATH" = c("kegg","path_id"),
                     "PMID" = c("pubmed","pubmed_id"),
                     "REFSEQ" = c("refseq","accession"),
                     "SYMBOL" = c("gene_info","symbol"),
                     "UNIGENE" = c("unigene","unigene_id"),
                     "ENSEMBL" = c("ensembl","ensembl_id"),
                     "ENSEMBLPROT" = c("ensembl_prot","prot_id"),
                     "ENSEMBLTRANS" = c("ensembl_trans","trans_id"),
                     "GENENAME" = c("gene_info","gene_name"),
                     "UNIPROT" = c("uniprot","uniprot_id"),
                     "GO" = c("go","go_id"),
                     "EVIDENCE" = c("go","evidence"),
                     "ONTOLOGY" = c("go","ontology"),
                     "GOALL" = c("go_all","go_id"),
                     "EVIDENCEALL" = c("go_all","evidence"),
                     "ONTOLOGYALL" = c("go_all","ontology")                     
                     )
  ## exceptions for ALL OrgDbs 
  if(class=="OrgDb"){
    ## I should probably remove ucsckg from select...
    #.defTables <- c(.defTables, list("UCSCKG" = c("ucsc","ucsc_id")) )
    .defTables <- .defTables[!(names(.defTables) %in% c("ALIAS2PROBE"))]
  }
  ## exceptions for ALL ChipDbs
  if(class=="ChipDb"){
    .defTables <- c(.defTables, list("PROBEID" = c("c.probes","probe_id")) )
  }

  ## species specific exceptions
  if(species=="Anopheles gambiae"){
    ## drop unsupported mappings
    .defTables <- .defTables[!(names(.defTables) %in% c("ALIAS",
                                                        "ALIAS2PROBE",
                                                        "MAP",
                                                        "CHRLOC",
                                                        "CHRLOCEND",
                                                        "CHRLOCCHR",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE") )]
  }
  if(species=="Arabidopsis thaliana"){
    ## add these
    .defTables <- c(.defTables, list("TAIR" = c("genes","gene_id"),
                                     "ARACYC" = c("aracyc","pathway_name"),
                                     "ARACYCENZYME" = c("enzyme","ec_name")))
    ## remove these:
    .defTables <- .defTables[!(names(.defTables) %in% c("ACCNUM",
                                                        "ALIAS",
                                                        "ALIAS2EG",
                                                        "ALIAS2PROBE",
                                                        "MAP",
                                                        "UNIGENE",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE",
                                                        "ENSEMBL",
                                                        "ENSEMBLPROT",
                                                        "ENSEMBLTRANS",
                                                        "UNIPROT",
                                                        "ENTREZID",
                                                        "CHR") )]
    ## "re-add" these (redefine, so must have been removed in prior step)
    .defTables <- c(.defTables, list("ENTREZID" = c("entrez_genes","gene_id"),
                                     "CHR"=c("gene_info","chromosome") ))
  }
  if(species=="Bos taurus"){
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP") )]
  }
  if(species=="Caenorhabditis elegans"){
    .defTables <- c(.defTables, list("WORMBASE" = c("wormbase","wormbase_id")))
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE") )]
  }
  if(species=="Canis familiaris"){
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE") )]
  }
  if(species=="Drosophila melanogaster"){
    .defTables <- c(.defTables, list("FLYBASE" = c("flybase","flybase_id"),
                                  "FLYBASECG" = c("flybase_cg","flybase_cg_id"),
                                  "FLYBASEPROT" = c("flybase_prot","prot_id")))
    .defTables <- .defTables[!(names(.defTables) %in% c("PFAM",
                                                        "IPI",
                                                        "PROSITE") )]
  }
  if(species=="Danio rerio"){
    .defTables <- c(.defTables, list("ZFIN" = c("zfin","zfin_id")))
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP"))]
  }
  if(species=="Escherichia coli"){
    .defTables <- .defTables[!(names(.defTables) %in% c("CHR",
                                                        "MAP",
                                                        "UNIGENE",
                                                        "CHRLOC",
                                                        "CHRLOCEND",
                                                        "CHRLOCCHR",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE",
                                                        "ENSEMBL",
                                                        "ENSEMBLPROT",
                                                        "ENSEMBLTRANS",
                                                        "UNIPROT"))]
  }
  if(species=="Gallus gallus"){
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP"))]
  }
  if(species=="Homo sapiens"){
    .defTables <- c(.defTables, list("OMIM" = c("omim","omim_id"),
                                     "UCSCKG" = c("ucsc","ucsc_id")) )
  }
  if(species=="Mus musculus"){
    .defTables <- c(.defTables, list("MGI" = c("mgi","mgi_id")) )#,
                                    # "UCSCKG" = c("ucsc","ucsc_id")) )
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP") )]
  }
  if(species=="Macaca mulatta"){
    .defTables <- .defTables[!(names(.defTables) %in% c("ALIAS",
                                                        "ALIAS2PROBE",
                                                        "MAP",
                                                        "UNIGENE",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE"))]
  }
  if(species=="Plasmodium falciparum"){
    .defTables <- c(.defTables, list(
      "ORF" = c("genes","gene_id") ))
    .defTables <- .defTables[!(names(.defTables) %in% c("ENTREZID",
                                                        "ACCNUM",
                                                        "ALIAS",
                                                        "ALIAS2PROBE",
                                                        "ALIAS2EG",
                                                        "CHR",
                                                        "CHRLOC",
                                                        "CHRLOCEND",
                                                        "CHRLOCCHR",
                                                        "MAP",
                                                        "PMID",
                                                        "REFSEQ",
                                                        "UNIGENE",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE",
                                                        "ENSEMBL",
                                                        "ENSEMBLPROT",
                                                        "ENSEMBLTRANS",
                                                        "UNIPROT") )]
    .defTables <- c(.defTables, list("ALIAS" = c("alias","alias_symbol") ))
  }
  if(species=="Pan troglodytes"){
    .defTables <- .defTables[!(names(.defTables) %in% c("ALIAS",
                                                        "ALIAS2PROBE",
                                                        "MAP",
                                                        "UNIGENE",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE") )]
  }
  if(species=="Rattus norvegicus"){
    #.defTables <- .defTables ## no changes (for now)
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP") )]
  }  
  if(species=="Saccharomyces cerevisiae"){
    .defTables <- c(.defTables, list(       
      "ORF" = c("gene2systematic","systematic_name"),
      "DESCRIPTION" = c("chromosome_features","feature_description"),
      "COMMON" = c("gene2systematic","gene_name"),
      "INTERPRO" = c("interpro","interpro_id"),
      "SMART" = c("smart","smart_id"),
      "SGD" = c("sgd","sgd_id") ))
    .defTables <- .defTables[!(names(.defTables) %in% c("ACCNUM",
                                                        "MAP",
                                                        "SYMBOL",
                                                        "UNIGENE",
                                                        "PROSITE",
                                                        "ALIAS",
                                                        "ALIAS2EG",
                                                        "ALIAS2PROBE",
                                                        "CHRLOC",
                                                        "CHRLOCEND",
                                                        "CHRLOCCHR",
                                                        "GENENAME",
                                                        "IPI",
                                                        "CHR") )]
    .defTables <- c(.defTables, list("ALIAS" = c("gene2alias","alias"),
                           "CHRLOC" = c("chromosome_features","start"),
                           "CHRLOCEND" = c("chromosome_features","stop"),
                           "CHRLOCCHR" = c("chromosome_features","chromosome"),
                           "GENENAME" = c("sgd","gene_name"),
                           "CHR" = c("chromosome_features","chromosome") ))
  }
  if(species=="Sus scrofa"){
    .defTables <- .defTables[!(names(.defTables) %in% c("MAP",
                                                        "CHRLOC",
                                                        "CHRLOCEND",
                                                        "CHRLOCCHR",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE",
                                                        "ENSEMBL",
                                                        "ENSEMBLPROT",
                                                        "ENSEMBLTRANS") )]
  }
  if(species=="Xenopus laevis"){
    .defTables <- .defTables[!(names(.defTables) %in% c("ALIAS",
                                                        "ALIAS2PROBE",
                                                        "MAP",
                                                        "CHRLOC",
                                                        "CHRLOCEND",
                                                        "CHRLOCCHR",
                                                        "PFAM",
                                                        "IPI",
                                                        "PROSITE",
                                                        "ENSEMBL",
                                                        "ENSEMBLPROT",
                                                        "ENSEMBLTRANS") )]
  }

  stockSpecies <- c("Anopheles gambiae",
                    "Arabidopsis thaliana",
                    "Bos taurus",
                    "Caenorhabditis elegans",
                    "Canis familiaris",
                    "Drosophila melanogaster",
                    "Danio rerio",
                    "Escherichia coli",
                    "Gallus gallus",
                    "Homo sapiens",
                    "Mus musculus",
                    "Macaca mulatta",
                    "Plasmodium falciparum",
                    "Pan troglodytes",
                    "Rattus norvegicus",
                    "Saccharomyces cerevisiae",
                    "Sus scrofa",
                    "Xenopus laevis")
  
  if(!(species %in% stockSpecies)){
    ## What follows is a very optimistic list!
  .defTables <- list("ENTREZID" = c("genes","gene_id"),
                     "ACCNUM" = c("accessions","accession"),
                     "ALIAS" = c("alias","alias_symbol"),
                     "ALIAS2EG" = c("alias","alias_symbol"),
                     "ALIAS2PROBE" = c("alias","alias_symbol"),
                     "CHR" = c("chromosomes","chromosome"),
                     "PMID" = c("pubmed","pubmed_id"),
                     "REFSEQ" = c("refseq","accession"),
                     "SYMBOL" = c("gene_info","symbol"),
                     "UNIGENE" = c("unigene","unigene_id"),
                     "GENENAME" = c("gene_info","gene_name"),
                     "GO" = c("go","go_id"),
                     "EVIDENCE" = c("go","evidence"),
                     "ONTOLOGY" = c("go","ontology")
                     )
  }

  
  ## ultimately I think I need GO.db to have it's OWN select methods.
  ## it's just too many responsibilities for this select to also handle the
  ## somewhat obscure GO database schema.  So many things will not really work
  ## untill I do that change (ancestors etc.)
  if(class=="GODb"){
    .defTables <- list("GOID" = c("go_term","go_id"),
                          "TERM" = c("go_term","term"),
                          "ONTOLOGY" = c("go_term","ontology"),
                          "DEFINITION" = c("go_term","definition")
                    ## "BPPARENTS" = c("go_bp_parents","_parent_id"),
                    ## "CCPARENTS" = c("go_cc_parents","_parent_id"),
                    ## "MFPARENTS" = c("go_mf_parents","_parent_id"),
                    ## "BPANCESTOR" = c("",""),
                    ## "CCANCESTOR" = c("",""),
                    ## "MFANCESTOR" = c("",""),
                    ## "OBSOLETE" = c("",""),
                    ## "SYNONYM" = c("go_synonym","synonym"),
                    ## "BPCHILDREN" = c("",""),
                    ## "CCCHILDREN" = c("",""),
                    ## "MFCHILDREN" = c("",""),
                    ## "BPOFFSPRING" = c("go_bp_offspring","_offspring_id"),
                    ## "CCOFFSPRING" = c("go_cc_offspring","_offspring_id"),
                    ## "MFOFFSPRING" = c("go_mf_offspring","_offspring_id")
                       )
  }

  
  ## then return with this result
  .defTables
}


## helper for getting out the table OR the field from .defineTables
.getDBLoc <- function(x, col, value="table"){
  res <- .defineTables(x)
  if(col %in% names(res)){
    res <- res[[col]]
  } else {
      stop("'col' value '", col, "' is not defined")
  }
  ## Then test and return appropriate records.
  if(value=="table"){
    res <- res[1]
  }else if(value=="field"){
    res <- res[2]
  }else if(value=="full.field"){
    res <- paste(res[1],res[2],sep=".")
  }
  res
}


## vectorized version of .getDBLoc
.getDBLocs <- function(x, cols, value="table"){
  res <- character(length(cols))
  for(i in seq_len(length(cols))){
      res[i] <- .getDBLoc(x, cols[i], value=value)
  }
  names(res) <- cols 
  unique(res)
}


## .getFullyQualifiedDBLocs <- function(x, cols){
##   tables <- .getDBLocs(x, cols, value="table")
##   fields <- .getDBLocs(x, cols, value="field")
##   paste(tables, fields, sep=".")
## }


## x is the org package object, y is the chip package object. 
.attachDB <- function(x,y){
  chipDb <- .getChipDbFile(y)
  chipSQL <- paste0("ATTACH '",chipDb,"' AS c")
  #message(chipSQL)
  dbAttach(dbconn(x), chipSQL)
}




.generateQuery <- function(x, cols, keytype, keys){
  ## If it is a chip package, get the org package instead
  if(class(x)=="ChipDb"){
    y <- x
    ## then flip to using the org package, and actually attach to that.
    x <- .getOrgPkg(x)
    try(.attachDB(x,y), silent=TRUE) ## not really a disaster if we fail here
    ## because we have a "y" defined, we have to define the dblocs this way:
    dblocs <- .getDBLocs(y, cols)
    ## Fully qualified keytype and species are set up like this
    ## fullKeytype <- .getFullyQualifiedDBLocs(y, keytype)
    fullKeytype <- .getDBLocs(y, keytype, value = "full.field")
    species <- species(x)
    ## if we have c.probes in dblocs, then we MUST join to genes table
    if("c.probes" %in% dblocs && species(x)!="Saccharomyces cerevisiae"){
      dblocs <- unique(append(dblocs, c("genes"), match("c.probes", dblocs)))
    }
    if("c.probes" %in% dblocs && species(x)=="Saccharomyces cerevisiae"){
      dblocs <- unique(append(dblocs, c("sgd"), match("c.probes", dblocs)))
    }
  }else{ ## this means there is only an org pkg...
    dblocs <- .getDBLocs(x, cols)
    ## fullKeytype <- .getFullyQualifiedDBLocs(x, keytype)
    fullKeytype <- .getDBLocs(x, keytype, value = "full.field")
    species <- species(x)
  }
  
  ## Get fields
  if(exists("y", inherits=FALSE)){ ## IOW if it was a chip package at the top...
    fields <- .getDBLocs(y,cols,value="full.field")
    fields <- paste(paste(fields,paste0(fields,"'"),sep=" AS '"), collapse=",")
  }else{
    fields <- .getDBLocs(x,cols,value="full.field")
    fields <- paste(paste(fields,paste0(fields,"'"),sep=" AS '"), collapse=",")
  }

  
  #message(paste(dblocs,collapse=","))
  ## then make the 1st part of the query.
  for(i in seq_len(length(dblocs))){
    if(i==1){
      res <- paste("SELECT ",fields," FROM",dblocs[i])
    }else{
##       if(species=="Saccharomyces cerevisiae" &&
##          (dblocs[i]=="gene2systematic" || dblocs[i-1]=="gene2systematic")){
##           join <- "systematic_name"
##       }else
      if(dblocs[i]=="c.probes" || dblocs[i-1]=="c.probes"){
         ## IOW if joining to OR from c.probes we want "gene_id"
        if(species=="Saccharomyces cerevisiae"){
          join <- "systematic_name"
        }else{
          join <- "gene_id"
        }
      }else{
        join <- "_id"
      }
      res <- c(res, paste("LEFT JOIN ",dblocs[i],"USING (",join,")"))
    }
  }
  res <- paste(res, collapse=" ")
  ## then use the keytype and keys to append the WHERE clause
  strKeys <- paste0('"',keys,'"',collapse=",")
  where <- paste("WHERE ",fullKeytype,"in (",strKeys,")" )
  paste(res, where)
}

## usage:
## library(org.Hs.eg.db)
## cols = c("ENTREZID","SYMBOL","PFAM")
## keytype = "ALIAS"
## keys = "ITGA7"
## x = org.Hs.eg.db
## AnnotationDbi:::.generateQuery(x, cols, keytype, keys)

## cols = c("ENTREZID","SYMBOL","PFAM")
## keytype = "ENTREZID"; keys = head(keys(org.Hs.eg.db)); x = org.Hs.eg.db
## AnnotationDbi:::.generateQuery(x, cols, keytype, keys)
## AnnotationDbi:::.extractData(x, cols, keytype, keys)
## select(org.Hs.eg.db,keys, cols, "ENTREZID")

## library(hgu95av2.db); x = hgu95av2.db;cols = c("ENTREZID","SYMBOL","PFAM")
## keytype = "PROBEID"; keys = head(keys(hgu95av2.db))
## AnnotationDbi:::.generateQuery(x, cols, keytype, keys) ## OK
## AnnotationDbi:::.extractData(x, cols, keytype, keys)
## select(hgu95av2.db,keys, cols, keytype)

## cols = c("ENTREZID", "SYMBOL", "CHRLOC")
## AnnotationDbi:::.extractData(x, cols, keytype, keys)
## select(hgu95av2.db,keys, cols, keytype)

## cols = c("ENTREZID", "SYMBOL", "GO")
## AnnotationDbi:::.extractData(x, cols, keytype, keys)
## select(hgu95av2.db,keys, cols, keytype)

## GO.db example attempt:
## library(GO.db);x<- GO.db;keys<-head(keys(GO.db)); cols = c("ONTOLOGY", "DEFINITION", "TERM"); keytype="GOID"
## AnnotationDbi:::.extractData(x, cols, keytype, keys)
## select(hgu95av2.db,keys, cols, keytype)


## I also need a method to call my generated sql and get the data.
## NOTE: the order that cols come back in is determined by the DB.
## Some resorting is done downstream.
.extractData <- function(x, cols, keytype, keys){
  ## Take the cols, append the keytype to FRONT
  cols <- unique(c(keytype, cols))
  ## do any necessary col expansion:
  cols <- unique(.expandCols(x, cols))
  ## generate the query
  sql <- .generateQuery(x, cols, keytype, keys)
  #message(sql)
  ## get field names for relevant cols
  cols <- unique(c(keytype, cols))
  headerTables <- .getDBLocs(x, cols, value="full.field")
  if(class(x)=="ChipDb"){
    y <- x ## save for test below
    x <- .getOrgPkg(x) ## then flip to using the org package
  }
  res <- dbQuery(dbconn(x), sql)
  ## then cleanup by doing a detach:
  if(exists("y", inherits=FALSE)){ ## I should not have to use inherits=FALSE?
    dbAttach(dbconn(x), "DETACH DATABASE c")
  }
  ## then subset to only relevant cols
  res[,(colnames(res) %in% headerTables)]
}


## usage:
## .extractData(x, cols, keytype, keys)



## helper used for dropping out ugly redundant col names.
## (BIMAP FREE!)
.simplifyCols <- function(x, cols){
  blackList <- c(ALIAS="ALIAS2PROBE",
                 ALIAS="ALIAS2EG",
                 CHR="CHRLOCCHR")
##                 CHR="CHRLOC",  ## Don't put cols here till removal is FINAL..
##                 CHR="CHRLOCEND",
##                 CHR="CHR")
  idx <- !(cols %in% blackList)
  unique(cols[idx])
}



.filterSuffixes <- function(tab){
  ## clean up .x and .y extensions?
  colnames(tab) <- gsub("\\.x","",colnames(tab))
  colnames(tab) <- gsub("\\.y","",colnames(tab))
  ## clean up .1's
  colnames(tab) <- gsub("\\.1","",colnames(tab))  
  tab
}



###############################################################################
## Helpers for tidying up the final table.
## resort_base drops unwanted rows, rearanges cols and puts things 
## into order that the keys were initially

## drop rows that don't match
.dropUnwantedRows <- function(tab, keys, jointype) {
    ## drop duplicated or 'all NA' (other than jointype) rows
    ntest <- ncol(tab) - sum(colnames(tab) == jointype)
    idx <- duplicated(tab) | (rowSums(is.na(tab)) == ntest)
    tab <- tab[!idx,, drop=FALSE]
    ## add back rows for keys that were completely removed
    noMatchKeys <- unique(keys[!keys %in% tab[[jointype]]])
    if (n <- length(noMatchKeys)) {
        ridx <- nrow(tab) + seq.int(n)
        cidx <- colnames(tab) %in% jointype
        tab[ridx, cidx] <- noMatchKeys
    }
    ## place rows in order of first appearance of key
    idx <- order(match(tab[[jointype]], keys))
    tab <- tab[idx,, drop=FALSE]
    rownames(tab) <- NULL
    tab
}

## resort the Column Names
.resortColumns <- function(tab, jointype, reqCols) {
    tab <- .filterSuffixes(tab) ## Removes duplicate suffixes
    if (!all(colnames(tab) %in% reqCols))
        stop("[internal] some of 'reqCols' are not in 'tab'")
    cnames <- c(jointype, reqCols[!reqCols %in% jointype])
    tab[, cnames, drop=FALSE]
}

## Create extra rows
.generateExtraRows <- function(tab, keys, jointype) {
    ## 4 possibilities
    ## if there are not dups, then we skip this function.
    ## if(any(duplicated(keys)) ## expand the keys
    ## if(any(duplicated(tab[[jointype]]))) ## some messages may apply
    ## AND if they are BOTH redundant how do I decide which row to expand?
    ## I think that I have to throw a warning and NOT do this step in that case?
    keyTest <- any(duplicated(keys))
    rowTest <-  any(duplicated(tab[[jointype]]))         
    if (keyTest && !rowTest){ 
        ind = match(keys, tab[[jointype]])
        tab <- tab[ind,,drop=FALSE]
        rownames(tab) <- NULL
    } else if (keyTest && rowTest) {
        indlst <- split(row.names(tab), tab[[jointype]])
        ind <- unlist(indlst[keys])
        tab <- tab[ind,]
    }
    ## We now just always give (terse) messages about relationship of
    ## data to columns returned.
    if(!keyTest && !rowTest) {
        txt <- "'select()' returned 1:1 mapping between keys and columns"  
    } else if (!keyTest && rowTest) {
        txt <- "'select()' returned 1:many mapping between keys and columns"
    } else if (keyTest && !rowTest) {
        txt <- "'select()' returned many:1 mapping between keys and columns"
    } else if (keyTest && rowTest) { ## User will get data "as is"
        txt <- "'select()' returned many:many mapping between keys and columns"
    }
    message(paste(strwrap(txt), collapse="\n"))
    tab
}

##  select() returned 1:1 mapping between keys and columns
##  select() returned 1:many mapping between keys and columns
##  select() returned many:1 mapping between keys and columns
## Also will want mesage for many:many

## .resort() was renamed resort_base()
## resort_base() is the main function for cleaning up a table so that 
## results look formatted the way we want them to.
resort_base <- function(tab, keys, jointype, reqCols) {
    if (jointype %in% colnames(tab)) {
        tab <- .dropUnwantedRows(tab, keys, jointype)
        ## rearrange to make sure cols are in correct order
        tab <- .resortColumns(tab, jointype, reqCols)
    }
    ## Duplicate any rows as appropriate (based on those keys)
    .generateExtraRows(tab, keys, jointype)
}



## helper so that we can be ready for when there are multiple things getting
## duplicated...
## The current version of these helper functions are not "smart", but it could
## be made so if needed.  The reason why not smart is because it will be a lot
## faster if we can get away with it being "dumb" and not looking up the types
## of values from the DB for each type.
.replaceValues <- function(dups, fieldNames, expectedCols){
  newVals <- expectedCols[fieldNames %in% dups]
  after <- match(dups, fieldNames) - 1 ## -1 b/c we aim to replac: not follow
  cols <- append(fieldNames, newVals, after)
  ## then remove the dups values
  cols[!(cols %in% dups)]
}


## helper for ambiguous/duplicated columns
.adjustForDupColNames <- function(res, expectedCols){
  fieldNames <- colnames(res)
  ## get duplicated vals
  dups <- fieldNames[duplicated(fieldNames)]
  ## for each value of dups, we want to call .replaceValues
  for(i in seq_len(length(dups))){
    if(i==1){
      cols <- .replaceValues(dups[i], fieldNames, expectedCols)
    }else{
      cols <- .replaceValues(dups[i], cols, expectedCols)
    }
  }
  cols
}


## helper looks at cols and trys to warn users about many:1 relationships
.warnAboutManyToOneRelationships <- function(cols){
    file=system.file("extdata","manyToOneBlackList.Rda",package="AnnotationDbi")
    blackList <- get(load(file))
    problemCols <- cols[cols %in% blackList]
    ## give message if there are more than 4 (for now)
    if(length(problemCols) > 4){
        msg = paste("You have selected the following columns that can have a many to one relationship with the primary key: ", paste(problemCols,collapse=", "),". Because you have selected more than a few such columns there is a risk that this selection may balloon up into a very large result as the number of rows returned multiplies accordingly. To experience smaller/more manageable results and faster retrieval times, you might want to consider selecting these columns separately.") 
        warning(paste(strwrap(msg, exdent=2), collapse="\n"),
                immediate.=TRUE, call.=FALSE)
    }
}


## the core of the select method for GO org and chip packages.
.legacySelect <- function(x, keys=NULL, cols=NULL, keytype, jointype) {
  ## IF CHR, CHR or CHRLOC are requested then you must deny the
  ## request as these are now deprecated
  if(any(.listDeprecatedKeytypes() %in% cols)){
      .deprecatedColsMessage()
  }
  ## if asked for what they have, just return that.
  if(all(cols %in% keytype)  && length(cols)==1){
    res <- data.frame(keys=keys)
    colnames(res) <- cols
    return(res)
  }
  if(is.null(keys)) keys <- keys(x) ## if no keys provided: use them all
  if(is.null(cols)) cols <- columns(x) ## if no cols provided: use them all
  
  ## call .simplifyCols to ensure we use same colnames as columns()
  cols <- .simplifyCols(x, cols)
  ## keytype <- .swapSymbolExceptions(x, keytype)
  keytype <- .simplifyCols(x, keytype)
  ## oriCols is a snapshot of col requests needed for column filter below
  oriCols <- unique(c(keytype, cols))


  ## Check if the user is selecting too many cols with many:1 relationships
  .warnAboutManyToOneRelationships(cols)

  ## Generate query and extract the data
  res <- .extractData(x, cols=cols, keytype=keytype, keys=keys)
  
  ## these are the colnames we need to have gotten back from the DB
  expectedCols <- .expandCols(x, oriCols)
  oriTabCols <- .getDBLocs(x, expectedCols, value="full.field")
  
  ## I need to know the jointype...
  jointype <- .getDBLocs(x, keytype, value="full.field")

  ## Remove suffixes in case there were dups
  res <- .filterSuffixes(res)

  
  ## If we can, then we should re-arrange to make sure cols come back in same
  ## order as they asked for initially.  Expanded cols cannot be re-arranged.
  if(all(expectedCols %in%  oriCols) &&
     any(oriCols != expectedCols) ){
    ## We need to make it so that oriTabCols is in the SAME order as oriCols
    oriTabCols <- .getDBLocs(x, oriCols, value="full.field")
    ## then we need to make expectedCols to match oriCols
    expectedCols <- oriCols
  }

  
  ## Then if any suffixes were actually removed, it means there were duplicated
  ## cols.  Duplicated cols means I have to do some label swapping.
  if( length(oriTabCols) < length(colnames(res))){
    oriTabCols <- .adjustForDupColNames(res, expectedCols) ## BADNESS!
    colnames(res) <- .adjustForDupColNames(res, expectedCols)
  }
  
  
  ## resort_base will resort the rows relative to the jointype etc.
  if(dim(res)[1]>0){
    res <- resort_base(res, keys, jointype, oriTabCols)
  }

  colnames(res) <- expectedCols[match(colnames(res), oriTabCols)]
  
  rownames(res) <- NULL
  res
}


## This just needs to generate a simple query and then return the
## results with no _id's
## A simple strategy will work EXCEPT for the weird case where I need
## to pull out multiple fields at once. (like GO)

## Do I need to pay attention to the order??? - it seems that I don't...
.appendGenesToTabs <- function(tabs){
    names <- names(tabs)
    tabs <- c("genes",tabs)
    names(tabs) <- c("GID", names)
    tabs
}

.noSchemaSelect <- function(x, keys=NULL, cols=NULL, keytype){

    ## 1st pool all the fields we need to extract
    fields <- unique(c(cols, keytype))
    ## Then get the tables to go with each one.
    tabs <- sapply(fields, .deriveTableNameFromField, x=x)
    ## make fully qualified fields of these tabs (the ones we want to extract)
    f.fields <- paste(tabs, fields, sep=".")

    ## if it's a Chip package, attach and point to org package for x
    if(class(x)=="ChipDb"){
        y <- x
        ## then flip to using the org package, and actually attach to that.
        x <- .getOrgPkg(x)
        try(.attachDB(x,y), silent=TRUE) ## not a disaster if we fail        
        ## Also make sure we include the genes table (only needed for the join)
        if(!("genes" %in% tabs)) tabs <- .appendGenesToTabs(tabs)
    }
    
    ## Make non-redundant list of tables to visit
    nrTabs <- unique(tabs)
    ## Now join to each table
    for(i in seq_along(nrTabs)){
        if(i==1){
            sql <- paste("SELECT ",paste(f.fields, collapse=","),
                         " FROM",tabs[1])
        }else{
            ## IF we see c.probes in nrTabs[i], it means we have to
            ## use gene_id instead.
            if("c.probes" %in% nrTabs[i]){
                sql <- c(sql, paste("LEFT JOIN ",nrTabs[i],"USING (GID)"))
            }else{
                sql <- c(sql, paste("LEFT JOIN ",nrTabs[i],"USING (_id)"))
            }
        }
    }
    sql <- paste(sql, collapse=" ")
    ## add the where clause
    strKeys <- paste0('"',keys,'"',collapse=",")
    fullKeytype <- tabs[names(tabs)==keytype]
    fullKeytype <- paste(fullKeytype, names(fullKeytype), sep=".") 
    where <- paste("WHERE ",fullKeytype,"in (",strKeys,")" )
    sql <- paste(sql, where)
    ## then call that
    res <- dbQuery(dbconn(x), sql)    
    ## cleanup and re-organize
    resort_base(res, keys, jointype=keytype, fields)
}

###############################################################################
## HELPERS for argument validation {in select() and keys()}

.isSingleString <- function(x){
  is.atomic(x) && length(x) == 1L && is.character(x)
}

testForValidKeytype <- function(x, keytype){
  if(!.isSingleString(keytype)){
      stop("'keytype' must be a a single string")
  }
  ## deprecated keytypes are still technically 'valid'
  pkts <- c(keytypes(x), .listDeprecatedKeytypes())
  if(!(keytype %in% pkts)){
      msg <- paste0("Invalid keytype: ",keytype,". Please use the keytypes method to see a listing of valid arguments.")
      stop(msg)
  }
}

.testForValidCols <- function(x, cols){
  if (!is.character(cols)){
      stop("'columns' must be a character vector")
  }
  pcols <- c(columns(x), .listDeprecatedKeytypes())
  if(!all(cols %in% pcols) && !is.null(cols)){
      badCols <- cols[!(cols %in% pcols)]
      msg <- paste0("Invalid columns: ",paste(badCols, collapse=","),". Please use the columns method to see a listing of valid arguments.")
      stop(msg)
  }
}

## fks is an alternate vector of keys to consult for validity.
## Normally this will be NULL and the test function should consult
## keys for the supplied keytype
.testForValidKeys <- function(x, keys, keytype, fks=NULL){
  if (!is.character(keys)){
      stop("'keys' must be a character vector")
  }
  if (length(keys) == 0L) {
      return()
  }
  if(is.null(fks)){  ## Normally, fks is just NULL and so we will call keys()
      ktKeys <- keys(x, keytype)
  }else{             ## This lets the caller say wait: use these keys instead 
      ktKeys <- fks 
  }
  if(!(any(ktKeys %in% keys))){
      msg <- paste0("None of the keys entered are valid keys for '",keytype,
         "'. Please use the keys method to see a listing of valid arguments.")
      stop(msg) ## later when things are better, demote this to a warning()
  }
}

testSelectArgs <- function(x, keys, cols, keytype, fks=NULL,
                            skipValidKeysTest=FALSE){
    testForValidKeytype(x, keytype)
    .testForValidCols(x, cols)
    ## I may need for this test to sometimes not happen...
    if(skipValidKeysTest==FALSE){
        .testForValidKeys(x, keys, keytype, fks)
    }
}
## library(Homo.sapiens); select(Homo.sapiens, c('11'), 'SYMBOL', 'GENEID')

## general select function
.select <- function(x, keys=NULL, cols=NULL, keytype, jointype, ...){
    ## Some argument handling and checking
    extraArgs <- list(...)
    if('fks' %in% names(extraArgs)){fks<-extraArgs[["fks"]]}else{fks<-NULL}
    if('skipValidKeysTest' %in% names(extraArgs)){
        skipValidKeysTest<-extraArgs[["skipValidKeysTest"]]}else{
            skipValidKeysTest<-FALSE}
    testSelectArgs(x, keys=keys, cols=cols, keytype=keytype,
           fks=fks, skipValidKeysTest=skipValidKeysTest)
    
    ## Now get the schema and call select
    schema <- metadata(x)[metadata(x)$name=="DBSCHEMA",]$value
    if(schema=="NOSCHEMA_DB" || schema=="NOCHIPSCHEMA_DB"){
        .noSchemaSelect(x, keys, cols, keytype)
    }else{
        .legacySelect(x, keys, cols, keytype, jointype)
    }
}




## Helper for setting the jointype to an appropriate default
.chooseJoinType  <- function(x){
  if(.getCentralID(x) == "ORF" && species(x) == "Saccharomyces cerevisiae"){
    jointype <- "gene2systematic.systematic_name"
  }else{
    jointype <- "genes.gene_id"
  }
  jointype
}


#######################################################################
## So an issue was that keytype was not defined carefully enough and
## was getting substituted when cols arg was used in conjunction with
## no keytype specified...  So when there is not keytype, the user
## must tell us which one it is OR we have to somehow "know"

## TO FIX: I need to guess the keytype and pass it along as some other
## name into ...
## Hack: I will pass it in as kt
## And halfway through this hack, it seems that if there is an
## argument passed in AFTER your other ones that the problem I was
## having magically clears up???  WTH?


setMethod("select", "OrgDb",
          function(x, keys, columns, keytype, ...) {
              if (missing(keytype)) keytype <- chooseCentralOrgPkgSymbol(x)
              jointype <- .chooseJoinType(x)
              .select(x, keys, columns, keytype, jointype=jointype, ...)
              ## .selectWarnJT(x, keys, columns, keytype, jointype=jointype,
              ##               kt=kt, ...)
          }
)

setMethod("select", "ChipDb",
    function(x, keys, columns, keytype, ...){
        if (missing(keytype)) keytype <- "PROBEID"
        .select(x, keys, columns, keytype, jointype="probes.probe_id", ...)
        ## .selectWarnJT(x, keys, columns, keytype, jointype="probes.probe_id",
        ##               kt=kt, ...)
    }
)

setMethod("select", "GODb",
    function(x, keys, columns, keytype, ...){
          if (missing(keytype)) keytype <- "GOID"
          .select(x, keys, columns, keytype, jointype="go_term.go_id", ...)
          ## .selectWarnJT(x, keys, columns, keytype, jointype="go_term.go_id",
          ## ...)
        }
)






#############################
## Internally we want to reconstruct these guys so we can merge() on them
## c <- columns(GO.db)[7]
## prefix = "GO"
## foo = paste0(prefix,c)
## bar = eval(parse(text=foo))



##############################################################################
## cols methods return the list of things that users can ask for.  This can be
## just the table names, or it might be a list of mappings

    
.legacyCols <- function(x, baseType){
  ## cols <- .makeColAbbrs(x)
  cols <- names(.defineTables(x))
  if(!missing(baseType)){
    cols <- c(baseType, cols)
  }
  ## translate relevant short bimap names to "cute" names
  cols <- .simplifyCols(x, cols)

  ## filter out columns that have been deprecated here
  hideCols <- .listDeprecatedKeytypes()
  cols <- cols[!(cols %in% hideCols)]
  
  ## .cols does not care about your names
  names(cols) <- NULL
  unique(cols)
}

## These helpers will go into the DB and extract col values for newer
## NOSCHEMA_DB's

.getDataTables <- function(con){
  tables <- dbListTables(con)
  tables[!tables %in% c("metadata","map_metadata","map_counts")]
}

.noSchemaCols <- function(x){
  if(class(x)=="ChipDb"){
      y <- x ## The old switcheroo
      x <- .getOrgPkg(x)  
  }
  con <- dbconn(x)
  tables <- .getDataTables(con)
  cols <- unique(unlist(sapply(tables, FUN=dbListFields, con=con)))
  cols <- cols[!cols %in% "_id"]
  if(exists('y', inherits=FALSE)){cols <- c("PROBEID", cols)}
  cols
}

## general .cols function
.cols <- function(x, baseType){
    schema <- metadata(x)[metadata(x)$name=="DBSCHEMA",]$value
    if(schema=="NOSCHEMA_DB" || schema=="NOCHIPSCHEMA_DB"){
        .noSchemaCols(x)
    }else{
        .legacyCols(x, baseType)
    }
}


setMethod("columns", "OrgDb",
    function(x){
      baseType <- chooseCentralOrgPkgSymbol(x)
      .cols(x, baseType)
    }
)

setMethod("columns", "ChipDb",
    function(x) .cols(x, baseType="PROBEID")
)

setMethod("columns", "GODb",
    function(x) .cols(x) ## does not have a missing baseType
)

#######################################################################
## Some testing of my deprecation:
## library(org.Hs.eg.db);  k = head(keys(org.Hs.eg.db, 'ENTREZID')); k; 
## head(keys(org.Hs.eg.db, 'CHR'));
## select(org.Hs.eg.db, k, 'SYMBOL', 'ENTREZID');
## head(select(org.Hs.eg.db, k[1], 'SYMBOL', 'CHR'));

## This method will just get all data for one column, one keytype 
## and one set of keys.  Then it will return either the 1st match for each, 
## filter out based on a rule OR return a CharacterList

mapIds_base <- function(x, keys, column, keytype, ..., multiVals=c("filter",
                        "asNA", "first","list","CharacterList")) {
    if (missing(multiVals)) 
        multiVals <- 'first'
    if (!is.function(multiVals))
        match.arg(multiVals)
    if (length(keys) < 1)
        stop(wmsg("mapIds must have at least one key to match against."))
    if (length(column) > 1)
        stop(wmsg("mapIds can only use one column."))
    if (length(keytype) >1 )
        stop(wmsg("mapIds can only use one keytype."))

    ## select, split and sort by keys
    res <- select(x, keys=unique(keys), columns=column, keytype=keytype)
    res <- split(res[[column]], f=res[[keytype]])[keys]

    ## handle multiple matches 
    .filter <- function(data) {
        idx <- elementNROWS(data) == 1
        nms <- names(data)[idx]
        data <- as.character(unlist(data[idx]))
        setNames(data, nms)
    }
    .asNA <- function(data) {
        idx <- elementNROWS(data) > 1
        data[idx] <- NA_character_
        unlist(data)
    }
    if (is.function(multiVals)) {
        sapply(res, FUN=multiVals)
    } else {
        switch(multiVals,
            "list" = res,
            "filter" = .filter(res),
            "asNA" = .asNA(res),
            "CharacterList" = as(res, 'CharacterList'),
            "first" = sapply(res, FUN=function(x) {x[[1]]})
        )
    }
}

setMethod("mapIds", "AnnotationDb", 
    function(x, keys, column, keytype, ..., multiVals)
        mapIds_base(x, keys, column, keytype, ..., multiVals=multiVals)
)

## TODO: add option to replace multi-matches with NAs or to just remove them.
## To cleanly handle having 'multiVals' being EITHER a FUN or something else:
## DO like: if(is.function(multiVals)){}else{match.arg(multiVals)}
## trace(mapIds, browser(), signature='AnnotationDb')
## library(org.Hs.eg.db); mapIds(org.Hs.eg.db, c('1','10'), 'ENTREZID', 'ENTREZID')


.taxonomyId <- function(x){
    conn <- dbconn(x)
    res <- dbGetQuery(conn,
                      'SELECT value from metadata where name like "TAXID" or name like "Taxonomy ID"')
    as.integer(res)
}

setMethod("taxonomyId", "AnnotationDb", function(x){.taxonomyId(x)})

Try the AnnotationDbi package in your browser

Any scripts or data that you put into this service are public.

AnnotationDbi documentation built on Nov. 8, 2020, 4:50 p.m.