diff --git a/r/DESCRIPTION b/r/DESCRIPTION
index ef461f78..8a921654 100644
--- a/r/DESCRIPTION
+++ b/r/DESCRIPTION
@@ -19,10 +19,13 @@ Suggests:
 Roxygen: list(markdown = TRUE)
 RoxygenNote: 6.1.1.9000
 Collate:
+    'read_funcs.R'
+    'nse_funcs.R'
+    'compile.R'
+    'datatable.R'
+    'entity.R'
+    'node.R'
     'stencila.R'
-    'util.R'
     'typing.R'
     'types.R'
-    'node.R'
-    'entity.R'
-    'datatable.R'
+    'util.R'
diff --git a/r/NAMESPACE b/r/NAMESPACE
index eeab1396..707fd1a4 100644
--- a/r/NAMESPACE
+++ b/r/NAMESPACE
@@ -2,28 +2,37 @@
 
 S3method(as.data.frame,Datatable)
 S3method(print,Entity)
+export(ArraySchema)
 export(Article)
 export(AudioObject)
 export(BlockContent)
+export(BooleanSchema)
 export(Brand)
+export(Cite)
+export(CiteGroup)
 export(Code)
 export(CodeBlock)
 export(CodeChunk)
-export(CodeExpr)
+export(CodeExpression)
+export(CodeFragment)
 export(Collection)
+export(ConstantSchema)
 export(ContactPoint)
 export(CreativeWork)
+export(CreativeWorkTypes)
 export(Datatable)
 export(DatatableColumn)
-export(DatatableColumnSchema)
 export(Delete)
 export(Emphasis)
 export(Entity)
+export(EnumSchema)
 export(Environment)
+export(Figure)
 export(Heading)
 export(ImageObject)
 export(Include)
 export(InlineContent)
+export(IntegerSchema)
 export(Link)
 export(List)
 export(ListItem)
@@ -31,16 +40,23 @@ export(Mark)
 export(MediaObject)
 export(Mount)
 export(Node)
+export(NumberSchema)
 export(Organization)
 export(Paragraph)
+export(Parameter)
+export(Periodical)
 export(Person)
 export(Product)
+export(PublicationIssue)
+export(PublicationVolume)
 export(Quote)
 export(QuoteBlock)
 export(ResourceParameters)
+export(SchemaTypes)
 export(SoftwareApplication)
 export(SoftwareSession)
 export(SoftwareSourceCode)
+export(StringSchema)
 export(Strong)
 export(Subscript)
 export(Superscript)
@@ -49,6 +65,7 @@ export(TableCell)
 export(TableRow)
 export(ThematicBreak)
 export(Thing)
+export(TupleSchema)
 export(VideoObject)
 export(as.Datatable.data.frame)
 export(datatable_from_dataframe)
@@ -56,3 +73,5 @@ export(datatable_to_dataframe)
 export(entity_from_list)
 export(node_from_json)
 export(node_to_json)
+export(nse_funcs)
+export(read_funcs)
diff --git a/r/R/compile.R b/r/R/compile.R
new file mode 100644
index 00000000..42ed389e
--- /dev/null
+++ b/r/R/compile.R
@@ -0,0 +1,161 @@
+#' @include nse_funcs.R
+#' @include read_funcs.R
+
+#' Names of function that assign
+assign_func_names <- c("assign", "base::assign", "<<-", "<-", "=")
+
+#' Names of functions that "import" packages
+import_func_names <- c("library", "require", "::", ":::")
+
+#' Names of function in the base R environment
+base_func_names <- ls(baseenv())
+
+# Temporary
+Function <- function(...) list(type = "Function", ...)
+
+compile_chunk <- function(chunk) {
+  language <- chunk$language
+  text <- chunk$text
+  imports <- NULL
+  declares <- NULL
+  assigns <- NULL
+  alters <- NULL
+  uses <- NULL
+  reads <- NULL
+
+  # Only handle R code
+  if (is.null(language) || !(language %in% c("r", "R"))) return(chunk)
+
+  # Parse the code into an AST
+  ast <- as.list(parse(text = text))
+
+  # Record assignments that are local
+  # to functions, they need to be considered
+  # for `uses`, but not for `assigns`
+  local_assigns <- NULL
+
+  ast_walker <- function(node, depth = 0) {
+    if (is.symbol(node)) {
+      name <- as.character(node)
+      if (!(name %in% assigns)) uses <<- unique(c(uses, name))
+    } else if (is.call(node)) {
+      # Resolve the function name
+      func <- node[[1]]
+      if (is.symbol(func)) {
+        # 'Normal' function call
+        # Add function to `uses` if it is is not in base environment
+        func_name <- as.character(func)
+        if (!(func_name %in% base_func_names)) {
+          uses <<- unique(c(uses, func_name))
+        }
+      } else if (is.call(func) && func[[1]] == "::") {
+        # Call of namespaced function e.g pkg::func
+        # Do not add these to `uses`
+        func_name <- paste0(func[[2]], "::", func[[3]])
+      } else {
+        # No func_name for other more complex calls
+        # that do not need to be detected below e.g. instance$method()
+        func_name <- ""
+      }
+
+      if (func_name == "$") {
+        # Only walk the left side, not the right since they are symbols to
+        # extract from an object so should not be included in `uses`
+        ast_walker(node[[2]], depth)
+        return()
+      } else if (func_name == "function") {
+        # Function definition
+        # Walk the body with incremented depth
+        ast_walker(node[[length(node)]], depth + 1)
+        return()
+      } else if (func_name %in% assign_func_names) {
+        left <- node[[2]]
+        right <- node[[3]]
+        if (is.call(right) && right[[1]] == "function") {
+          # Assignment of a function
+          # Treat as a declaration
+          func_decl <- Function(
+            name = as.character(left)
+          )
+          if (!is.null(right[[2]])) {
+            parameters <- NULL
+            params <- as.list(right[[2]])
+            print(names(params))
+            for (name in names(params)) {
+              param <- params[[name]]
+              parameters <- c(parameters, list(
+                Parameter(
+                  name = name
+                )
+              ))
+            }
+            func_decl$parameters <- parameters
+          }
+          declares <<- c(declares, list(func_decl))
+
+          # Only walk the function so that left is not made a `uses`
+          ast_walker(right, depth)
+          return()
+        } else if (func == "assign") {
+          # Assignment using `assign` function
+          # TODO: Check the `pos` arg relative to current depth
+          assigns <<- unique(c(assigns, left))
+        } else if (is.symbol(left) && depth == 0) {
+          # Assignment to a name
+          assigns <<- unique(c(assigns, as.character(left)))
+        } else if (is.call(left)) {
+          # Assignment to an existing object  e.g. a$b[1] <- NULL
+          # Recurse until we find the variable that is target of alteration
+          walk <- function(node) {
+            target <- node[[2]]
+            if (is.symbol(target)) {
+              if (is.null(alters) || !(target %in% alters)) {
+                alters <<- c(alters, as.character(target))
+              }
+            } else {
+              walk(target)
+            }
+          }
+          walk(target)
+        }
+      } else if (func_name %in% import_func_names) {
+        # Package import
+        # Get the names of the package
+        if (length(node) > 1) {
+          imports <<- unique(c(imports, as.character(node[[2]])))
+        }
+      } else if (func_name %in% read_funcs_names) {
+        # File read
+        # Collect relevant argument(s) from function call
+        args <- as.list(node[2:length(node)])
+        read_func_index <- floor((match(func_name, read_funcs_names) - 1) / 2) + 1
+        read_func <- read_funcs[[read_func_index]]
+        if (any(read_func$names %in% names(args))) {
+          files <- unlist(args[read_func$names])
+        } else {
+          files <- unlist(args[read_func$positions])
+        }
+        # Only use character arguments i.e. not symbols (variable names)
+        files <- files[is.character(files)]
+        if (length(files) > 0) reads <<- unique(c(reads, files))
+      }
+    }
+
+    # If there are child nodes, walk over them too
+    if (length(node) > 1) {
+      lapply(node[2:length(node)], ast_walker, depth)
+    }
+  }
+  lapply(ast, ast_walker)
+
+  list(
+    language = language,
+    text = text,
+    imports = imports,
+    declares = declares,
+    assigns = assigns,
+    alters = alters,
+    uses = uses,
+    reads = reads
+  )
+}
diff --git a/r/R/nse_funcs.R b/r/R/nse_funcs.R
new file mode 100644
index 00000000..46e7856f
--- /dev/null
+++ b/r/R/nse_funcs.R
@@ -0,0 +1,30 @@
+#' Module for defining a list of non-standard evaluation functions,
+#' functions that use `substitute()` (or related) on one or more arguments.
+#' See http://adv-r.had.co.nz/Computing-on-the-language.html.
+#' This list is used when R code is compiled to ignore some
+#' variable names when determining the `uses` property
+#' of the chunk.
+
+#' Create a entry for a function that uses NSE
+#'
+#' @param func Name of the function
+#' @param package Name of the package that the function is in
+#' @param names Names of parameters that should be ignored
+#' @param positions Positions of parameters that should be ignored
+nse_func <- function(func, package, names=NULL, positions=NULL) {
+  list(
+    func = func,
+    package = package,
+    names = names,
+    positions = positions
+  )
+}
+
+#' List of functions that read from files
+#' @export
+nse_funcs <- list(
+  # base package
+  nse_func("base", "subset", c("subset", "select"), c(2, 4)),
+  # dplyr package
+  nse_func("dplyr", "filter")
+)
diff --git a/r/R/read_funcs.R b/r/R/read_funcs.R
new file mode 100644
index 00000000..ab76ab81
--- /dev/null
+++ b/r/R/read_funcs.R
@@ -0,0 +1,54 @@
+#' Module for defining a list of "read" functions, functions
+#' that read files from the filesystem. This list is used
+#' when R code is compiled to help determine the `reads` propery
+#' of the chunk
+
+#' Create a entry for a function that reads a file
+#'
+#' Most file reading functions have the file path as their first
+#' parameter named `file`. If this is not the case, or if there
+#' is more than one parameter that relates to a file that is read
+#' by the function, use the `names` and `positions` parameters
+#'
+#' @param package Name of the package that the function is in
+#' @param func Name of the function
+#' @param names Names of parameters that are file paths that are read
+#' @param positions Positions of parameters that are file paths that are read
+read_func <- function(package, func, names="file", positions=1) {
+  list(
+    package = package,
+    func = func,
+    names = names,
+    positions = positions
+  )
+}
+
+#' List of functions that read from files
+#' @export
+read_funcs <- list(
+  # utils package
+  read_func("utils", "read.table"),
+  read_func("utils", "read.csv"),
+  read_func("utils", "read.csv2"),
+  read_func("utils", "read.delim"),
+  read_func("utils", "read.delim2"),
+  read_func("utils", "read.fwf"),
+  # foreign package
+  read_func("foreign", "read.arff"),
+  read_func("foreign", "read.dbf"),
+  read_func("foreign", "read.dta"),
+  read_func("foreign", "read.epiinfo"),
+  read_func("foreign", "read.mtp"),
+  read_func("foreign", "read.octave"),
+  read_func("foreign", "read.spss"),
+  read_func("foreign", "read.ssd"),
+  read_func("foreign", "read.systat"),
+  read_func("foreign", "read.xport")
+)
+
+#' List of possible function call names to match
+read_funcs_names <- Reduce(
+  function(prev, curr) c(prev, curr$func, paste0(curr$package, "::", curr$func)),
+  read_funcs,
+  character()
+)
diff --git a/r/R/types.R b/r/R/types.R
index 4a739fae..77f5e6a3 100644
--- a/r/R/types.R
+++ b/r/R/types.R
@@ -2,6 +2,8 @@
 # Do not modify it by hand. Instead, modify the source `.schema.yaml` files
 # in the `schema` directory and run `npm run build:r` to regenerate it.
 
+#' @include typing.R
+
 #' The most basic item, defining the minimum properties required.
 #'
 #' @name Entity
diff --git a/r/R/util.R b/r/R/util.R
index da3976a0..434fb88f 100644
--- a/r/R/util.R
+++ b/r/R/util.R
@@ -1,25 +1,26 @@
 # Utility functions used internally in this package
 # and not intended to be exported.
 
-#' Map a function across entries in a node
+#' Map a function across entries in an object
 #'
-#' This is analagous to `Object.entries(node).map(...)`
+#' This is analagous to `Object.entries(object).map(...)`
 #' in Javascript. It handles bother scalar and vector
-#' node types.
+#' object types.
 #'
-#' @param node The node to map over
-#' @param func The function to apply to each of the node's entries
+#' @param object The object to map over
+#' @param func The function to apply to each of the object's entries
 #' @param ... Additional arguments to pass through to the function
-map <- function(node, func, ...) {
-  if (is.list(node)) lapply(node, func, ...)
-  else func(node, ...)
+map <- function(object, func, ...) {
+  if (is.list(object)) lapply(object, func, ...)
+  else func(object, ...)
 }
 
-#' Transform a node by recursively applying a function to it.
+#' Create a transformattion of a object by recursively
+#' applying a function to it. Could be called `deepMap`.
 #'
-#' @param node The node to map over
-#' @param func The function to apply to each node
+#' @param object The object to map over
+#' @param func The function to apply to each object
 #' @param ... Additional arguments to pass through to the function
-transform <- function(node, func, ...) {
-  map(node, function(child) map(child, func, ...))
+transform <- function(object, func, ...) {
+  map(object, function(child) map(child, func, ...))
 }