From e24fc95d46d1f4228df259f0885cf83680c51871 Mon Sep 17 00:00:00 2001
From: "Maarten A. Breddels" <maartenbreddels@gmail.com>
Date: Mon, 13 Jul 2020 12:59:47 +0200
Subject: [PATCH] core(fix): offsets for arrow string array not respected in
 conversion

---
 packages/vaex-core/vaex/arrow/convert.py | 3 ++-
 packages/vaex-core/vaex/dataframe.py     | 2 +-
 tests/arrow/conversion_test.py           | 9 +++++++++
 3 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/packages/vaex-core/vaex/arrow/convert.py b/packages/vaex-core/vaex/arrow/convert.py
index 34ddb0e59d..cbae35651c 100644
--- a/packages/vaex-core/vaex/arrow/convert.py
+++ b/packages/vaex-core/vaex/arrow/convert.py
@@ -93,7 +93,8 @@ def column_from_arrow_array(arrow_array):
             string_bytes = np.array([], dtype='S1')
         else:
             string_bytes = np.frombuffer(string_bytes, 'S1', len(string_bytes))
-        column = vaex.column.ColumnStringArrow(offsets, string_bytes, len(arrow_array), null_bitmap=null_bitmap)
+        offset = arrow_array.offset
+        column = vaex.column.ColumnStringArrow(offsets[offset:], string_bytes, len(arrow_array), null_bitmap=null_bitmap)
         return column
     else:
         raise TypeError('type unsupported: %r' % arrow_type)
diff --git a/packages/vaex-core/vaex/dataframe.py b/packages/vaex-core/vaex/dataframe.py
index 433b43aac6..fbfcb2e5d2 100644
--- a/packages/vaex-core/vaex/dataframe.py
+++ b/packages/vaex-core/vaex/dataframe.py
@@ -501,7 +501,7 @@ def map(thread_index, i1, i2, ar):
                     if not transient:
                         assert ar is previous_ar.string_sequence
                 # TODO: what about masked values?
-                inverse[i1:i2:] = ordered_set.map_ordinal(ar)
+                inverse[i1:i2] = ordered_set.map_ordinal(ar)
             def reduce(a, b):
                 pass
             self.map_reduce(map, reduce, [expression], delay=delay, name='unique_return_inverse', info=True, to_numpy=False, selection=selection)
diff --git a/tests/arrow/conversion_test.py b/tests/arrow/conversion_test.py
index 99ee1ade04..08d7acfb14 100644
--- a/tests/arrow/conversion_test.py
+++ b/tests/arrow/conversion_test.py
@@ -2,6 +2,7 @@
 import pyarrow as pa
 import pytest
 from vaex import array_types 
+import vaex.arrow.convert
 
 
 bools = [False, True, True]
@@ -48,6 +49,14 @@ def test_float_sliced_masked():
     assert x.tolist() == x_original[2:].tolist()
 
 
+def test_string_sliced():
+    values = ["a", "bb", "ccc", "dddd"]
+    ar = pa.array(values)
+    assert vaex.arrow.convert.column_from_arrow_array(ar).tolist() == values
+    assert vaex.arrow.convert.column_from_arrow_array(ar[1:]).tolist() == values[1:]
+    assert vaex.arrow.convert.column_from_arrow_array(ar[1:3]).tolist() == values[1:3] != values[1:]
+
+
 def test_keep_masked_data_values():
     x_original = np.arange(5, dtype='f8')
     mask = x_original > 2