[improve] Treat “classful” properties separately

This lets the `stopifnot` check become bi-directional.

Author: Dominique Quatravaux

confluence.R

@@ -38,17 +38,18 @@ confluence_id <- function (node_or_nodeset) {
         xml_text
 }
 
-#' Format the “props” of each node in a nodeset, into a tibble
+#' Format the classless “props” of each node in a nodeset, into a tibble
 #'
 #' @param objects_nodeset An XML nodeset (*not* a list of many
 #'     nodesets)
 #' @return A tibble with as many rows as there are objects in
 #'     objects_nodeset. The columns of the returned tibble are the XML
-#'     “name” attributes of any property nodes found, and the values
-#'     are the XML texts found within said property nodes.
+#'     “name” attributes of any `property` sub-nodes that do *not*
+#'     have a `class` attribute; and the values are the XML texts
+#'     found within said property nodes.
 props_tibble <- function(objects_nodeset) {
     objects_nodeset %>%
-    xml_find_all('property', flatten=FALSE) %>%
+    xml_find_all('property[not(@class)]', flatten=FALSE) %>%
     tibble(row = seq_along(.),
            props = .) %>%
         rowwise() %>%
@@ -60,6 +61,32 @@ props_tibble <- function(objects_nodeset) {
         select(-row)
 }
 
+#' Format the “props” that have a `class` for each node in a nodeset, into a tibble
+#'
+#' @param objects_nodeset An XML nodeset (*not* a list of many
+#'     nodesets)
+#' @return A tibble with as many rows as there are objects in
+#'     objects_nodeset. The columns of the returned tibble are named
+#'     as `foo.Bar`, where `foo` (e.g. "content") is the `name`
+#'     attribute, and `Bar` is the `class` attribute, of any
+#'     `property` sub-nodes that have a `class` attribute; and the
+#'     values are the XML texts found within said property nodes.
+classful_props_tibble <- function(objects_nodeset) {
+    objects_nodeset %>%
+    xml_find_all('property[@class]', flatten=FALSE) %>%
+    tibble(row = seq_along(.),
+           props = .) %>%
+        rowwise() %>%
+        reframe(
+            row = row,
+            pnames = xml_attr(props, 'name'),
+            pclasses = xml_attr(props, 'class'),
+            pvals = xml_text(props)) %>%
+        mutate(pkeys=paste(pnames, pclasses, sep="."), .keep = "unused") %>%
+        pivot_wider(names_from = pkeys, values_from = pvals) %>%
+        select(-row)
+}
+
 ## Again, stuff like
 ##
 ##      object_pages <- entities %>%
@@ -74,6 +101,7 @@ page_versions <- {
     ns <- entities_xml %>% xml_find_all('//object[@class="Page"]')
     tibble(content_id = ns %>% confluence_id) %>%
         mutate(ns %>% props_tibble) %>%
+        mutate(ns %>% classful_props_tibble) %>%
         mutate(ns %>%
                xml_find_all(c('collection[@name="contentProperties"]',
                               'element[@class="ContentProperty"]',
@@ -88,21 +116,22 @@ content_properties <- {
     ns <- entities_xml %>% xml_find_all('//object[@class="ContentProperty"]')
     tibble(property_id = ns %>% confluence_id) %>%
         mutate(ns %>% props_tibble) %>%
-        rename(content_id = content)
+        mutate(ns %>% classful_props_tibble)
 }
 
 ## page_versions$content_property_ids (as a “multivalued foreign key”)
-## ought to be a subset of content_properties[c("content.Page",
-## "property_id")]:
+## ought to contain the same information as
+## content_properties[c("content.Page", "property_id")]:
 stopifnot({
     relation1 <-
         page_versions %>%
         transmute(content_id, property_id = content_property_ids) %>%
         unnest_longer(property_id)
     relation2 <-
         content_properties %>%
+        filter(! is.na(content.Page)) %>%
         transmute(content_id = content.Page, property_id)
-    ## The opposite is not true, because some properties are for objects
-    ## other than pages:
-    anti_join(relation1, relation2, by = join_by(content_id, property_id)) %>% nrow == 0
+    by <- join_by(content_id, property_id)
+    anti_join(relation1, relation2, by = by) %>% nrow == 0 &&
+        anti_join(relation2, relation1, by = by) %>% nrow == 0
 })