diff --git a/README.md b/README.md
index c891b56..4ab05dc 100644
--- a/README.md
+++ b/README.md
@@ -307,12 +307,14 @@ user> (classifier-classify classifier2 instance)
 Text document handling:
 
 ```clojure
-user> (def docs [{:title "Document title 1"
+user> (def docs [{:id 10
+                  :title "Document title 1"
                   :fulltext "This is the fulltext..."
-                  :has-term? false}
-                 {:title "Another document title"
+                  :has-class? false}
+                 {:id 11
+                  :title "Another document title"
                   :fulltext "Some more \"fulltext\"; rabbit artificial machine bananas"
-                  :has-term? true}])
+                  :has-class? true}])
 #'user/docs
 
 user> (docs-to-dataset docs "bananas-model" "my-models" :stemmer true :lowercase false)
diff --git a/src/clj_ml/data.clj b/src/clj_ml/data.clj
index 6deb6ef..f701e66 100644
--- a/src/clj_ml/data.clj
+++ b/src/clj_ml/data.clj
@@ -535,7 +535,7 @@ split immediately you can use do-split-dataset."
   [docs model-prefix model-dir & opts]
   (let [parsed-opts (apply hash-map opts)
         original-ordering (map :id docs)
-        docs-with-class (filter :has-class? doc)
+        docs-with-class (filter :has-class? docs)
         docs-without-class (let [dwoc (filter #(not (:has-class? %)) docs)]
                              (if (:resample parsed-opts)
                                (take (count docs-with-class) dwoc)
@@ -544,7 +544,7 @@ split immediately you can use do-split-dataset."
                       (concat (take (/ (:keep-n parsed-opts) 2) docs-with-class)
                               (take (/ (:keep-n parsed-opts) 2) docs-without-class))
                       (concat docs-with-class docs-without-class))
-        docs-shuffled (my-shuffle (sort-by :id docs-keep-n))
+        docs-shuffled (shuffle (sort-by :id docs-keep-n))
         ds (make-dataset
             :docs [{:class [:no :yes]} {:title nil} {:fulltext nil}]
             (for [doc docs-shuffled]
diff --git a/test/clj_ml/data_test.clj b/test/clj_ml/data_test.clj
index b7beb17..d4ce413 100644
--- a/test/clj_ml/data_test.clj
+++ b/test/clj_ml/data_test.clj
@@ -229,3 +229,20 @@
                           [3 :bar]
                           [4 :foo]])]
     (is (= (dataset-as-vecs (take-dataset ds 2)) [[1.0 :foo] [2.0 :bar]]))))
+
+(deftest docs-to-dataset-test
+  (let [docs [{:id 10
+               :title "Document title 1"
+               :fulltext "This is the fulltext..."
+               :has-class? false}
+              {:id 11
+               :title "Another document title"
+               :fulltext "Some more \"fulltext\"; rabbit artificial machine bananas"
+               :has-class? true}]
+        {:keys [dataset docids]} (docs-to-dataset docs "bananas-model" "."
+                                                  :stemmer true :lowercase false)
+        docid-ds-vecs (apply hash-map (interleave docids (dataset-as-vecs dataset)))]
+    (is (= [:no 0.4804530139182014 0.0 0.4804530139182014 0.0]
+           (get docid-ds-vecs 10)))
+    (is (= [:yes 0.0 0.4804530139182014 0.0 0.4804530139182014 0.0 0.4804530139182014 0.0 0.4804530139182014]
+           (get docid-ds-vecs 11)))))