Merge pull request #13 from hawkfish/voila

Various fixes found working on the Voila example

src/textform/fold.py

@@ -70,6 +70,9 @@ def readrow(self):
         #   Buffer empty, so pivot next row
         row = super().readrow()
 
+        #   Rebuild the buffer because the rows we return are not ours.
+        self._buffer = [{self.tag: tag} for tag in self.tags]
+
         #   Update the folds
         for g, group in enumerate(self._groups):
             output = self.outputs[g+1]

src/textform/iterate.py

@@ -29,7 +29,7 @@ class Iterate(Transform):
     def __init__(self, source, input, tags, strings, layout='csv', **config):
         super().__init__('iterate', input, (tags, strings,), source)
 
-        self._validateOutputs()
+        self._validateOutputs(self.inputs)
 
         self.function = bind_iterate(self.name, layout, strings, **config)
         self._buffer = []
@@ -54,12 +54,13 @@ def readrow(self):
         if self._position < len(self._buffer):
             return self._unbuffer()
 
-        #   Buffer empty, so expand the next row
+        #   Buffer flushed, so expand the next row
         row = self.source.readrow()
         ragged = self.function(row[self.input])
         del row[self.input]
 
         #   Update the buffer
+        self._buffer = []
         for tag in ragged:
             buffered = dict(zip(self.outputs, (str(tag), str(ragged[tag]),)))
             buffered.update(row)

src/textform/unfold.py

@@ -15,6 +15,9 @@ def __init__(self, source, inputs, outputs):
 
         self.fixed = [f for f in filter(lambda input: input not in self.inputs, self.source.fieldnames)]
 
+        # Ragged group buffer
+        self._ragged = None
+
     def _schema(self):
         schema = super()._schema()
 
@@ -46,10 +49,28 @@ def readrow(self):
         #   and emitting a row whenever it is complete
         row = None
         for g in range(len(self._groups[0])):
-            folded = super().readrow()
-
-            if row is None: row = {fixed: folded[fixed] for fixed in self.fixed}
-            row.update({self._groups[f][g]: folded[self.folds[f]] for f in range(len(self.folds))})
+            folded = self._ragged
+            self._ragged = None
+            if not folded:
+                try:
+                    folded = super().readrow()
+                except StopIteration:
+                    if row: break
+                    raise
+
+            if row is None:
+                row = {output: None for output in self.outputs}
+                row.update({fixed: folded[fixed] for fixed in self.fixed})
+
+            elif not all([folded[fixed] == row[fixed] for fixed in self.fixed]):
+                self._ragged = {output: None for output in self.outputs}
+                self._ragged.update(folded)
+                break
+
+            try:
+                row.update({self._groups[f][g]: folded[self.folds[f]] for f in range(len(self.folds))})
+            except:
+                raise
 
         self._updateSchemaTypes(row, self.outputs)
 

tests/test_fold.py

@@ -129,3 +129,12 @@ def test_voila_6(self):
             self.assertEqual(5 + actual // len(unfolded), row['#BLENDs'], actual)
             actual += 1
         self.assertEqual((len(csv) - 1) * len(unfolded), actual)
+
+    def test_drop_tags(self):
+        s = None
+        s = txf.Sequence(s, 'Fold 1')
+        s = txf.Sequence(s, 'Fold 2')
+        s = txf.Limit(s, 4)
+        s = txf.Fold(s, ('Fold 1', 'Fold 2',), ('Tags', 'Group',))
+        s = txf.Drop(s, 'Tags')
+        s.pump()