X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=branch%2Fpatch%2FR%2Frdep.R;fp=branch%2Fpatch%2FR%2Frdep.R;h=141453b80b90ca929d17aab48b9b314e065bdce0;hb=75ac4a666a0faa4a2d09aeed4db94746c097b1ca;hp=0000000000000000000000000000000000000000;hpb=7c74d4e371b7ecda2035aa106918b0427818cacf;p=cran2deb.git

diff --git a/branch/patch/R/rdep.R b/branch/patch/R/rdep.R
new file mode 100644
index 0000000..141453b
--- /dev/null
+++ b/branch/patch/R/rdep.R
@@ -0,0 +1,155 @@
+
+r_bundle_of <- function(pkgname) {
+    # returns the bundle containing pkgname or NA
+    bundles <- names(available[!is.na(available[, 'Bundle']), 'Contains'])
+    # use the first bundle
+    for (bundle in bundles) {
+        if (pkgname %in% r_bundle_contains(bundle)) {
+            return(bundle)
+        }
+    }
+    return(NULL)
+}
+
+r_bundle_contains <- function(bundlename) {
+    return(strsplit(available[bundlename,'Contains'],'[[:space:]]+')[[1]])
+}
+
+r_requiring <- function(names) {
+    for (name in names) {
+        if (!(name %in% base_pkgs) && !(name %in% rownames(available))) {
+            bundle <- r_bundle_of(name)
+            if (!is.null(bundle)) {
+                name = bundle
+                names <- c(names,bundle)
+            }
+        }
+        if (name %in% rownames(available) && !is.na(available[name,'Contains'])) {
+            names <- c(names,r_bundle_contains(name))
+        }
+    }
+    # approximately prune first into a smaller availability
+    candidates <- rownames(available)[sapply(rownames(available)
+                                            ,function(name)
+                                                length(grep(paste(names,collapse='|')
+                                                           ,available[name,r_depend_fields])) > 0)]
+    if (length(candidates) == 0) {
+        return(c())
+    }
+    # find a logical index into available of every package/bundle
+    # whose dependency field contains at least one element of names.
+    # (this is not particularly easy to read---sorry---but is much faster than
+    # the alternatives i could think of)
+    prereq=c()
+    dep_matches <- function(dep) chomp(gsub('\\([^\\)]+\\)','',dep)) %in% names
+    any_dep_matches <- function(name,field=NA)
+                any(sapply(strsplit(chomp(available[name,field])
+                                   ,'[[:space:]]*,[[:space:]]*')
+                          ,dep_matches))
+
+    for (field in r_depend_fields) {
+        matches = sapply(candidates, any_dep_matches, field=field)
+        if (length(matches) > 0) {
+            prereq = c(prereq,candidates[matches])
+        }
+    }
+    return(unique(prereq))
+}
+
+r_dependencies_of <- function(name=NULL,description=NULL) {
+    # find the immediate dependencies (children in the dependency graph) of an
+    # R package
+    if (!is.null(name) && (name == 'R' || name %in% base_pkgs)) {
+        return(data.frame())
+    }
+    if (is.null(description) && is.null(name)) {
+        fail('must specify either a description or a name.')
+    }
+    if (is.null(description)) {
+        if (!(name %in% rownames(available))) {
+            bundle <- r_bundle_of(name)
+            if (!is.null(bundle)) {
+                name <- bundle
+            } else {
+                # unavailable packages don't depend upon anything
+                return(data.frame())
+            }
+        }
+        description <- data.frame()
+        # keep only the interesting fields
+        for (field in r_depend_fields) {
+            if (!(field %in% names(available[name,]))) {
+                next
+            }
+            description[1,field] = available[name,field]
+        }
+    }
+    # extract the dependencies from the description
+    deps <- data.frame()
+    for (field in r_depend_fields) {
+        if (!(field %in% names(description[1,]))) {
+            next
+        }
+        new_deps <- lapply(strsplit(chomp(description[1,field])
+                                   ,'[[:space:]]*,[[:space:]]*')[[1]]
+                          ,r_parse_dep_field)
+        deps <- iterate(lapply(new_deps[!is.na(new_deps)],rbind),deps,rbind)
+    }
+    return (deps)
+}
+
+r_parse_dep_field <- function(dep) {
+    if (is.na(dep)) {
+        return(NA)
+    }
+    # remove other comments
+    dep = gsub('(\\(\\)|\\([[:space:]]*[^<=>!].*\\))','',dep)
+    # squish spaces
+    dep = chomp(gsub('[[:space:]]+',' ',dep))
+    # parse version
+    pat = '^([^ ()]+) ?(\\( ?([<=>!]+ ?[0-9.-]+) ?\\))?$'
+    if (!length(grep(pat,dep))) {
+        fail('R dependency',dep,'does not appear to be well-formed')
+    }
+    version = sub(pat,'\\3',dep)
+    dep = sub(pat,'\\1',dep)
+    if (!(dep %in% rownames(available))) {
+        depb <- r_bundle_of(dep)
+        if (!is.null(depb)) {
+            dep <- depb
+        }
+    }
+    return(list(name=dep,version=version))
+}
+
+r_dependency_closure <- function(fringe, forward_arcs=T) {
+    # find the transitive closure of the dependencies/prerequisites of some R
+    # packages
+    closure <- list()
+    if (is.data.frame(fringe)) {
+        fringe <- as.list(fringe$name)
+    }
+    fun = function(x) r_dependencies_of(name=x)$name
+    if (!forward_arcs) {
+        fun = r_requiring
+    }
+    while(length(fringe) > 0) {
+        # pop off the top
+        top <- fringe[[1]]
+        if (length(fringe) > 1) {
+            fringe <- fringe[2:length(fringe)]
+        } else {
+            fringe <- list()
+        }
+        src <- pkgname_as_debian(top,binary=F)
+        if (src == 'R') {
+            next
+        }
+        newdeps <- fun(top)
+        closure=c(closure,top)
+        fringe=c(fringe,newdeps)
+    }
+    # build order
+    return(rev(unique(closure,fromLast=T)))
+}
+