From 7138a352a7075a31da6174951c609f434e50e59c Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Tue, 16 Jul 2024 16:40:21 -0400
Subject: [PATCH 01/42] Add new type to header

---
 gsd/gsd.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/gsd/gsd.h b/gsd/gsd.h
index 194b3750..388e4201 100644
--- a/gsd/gsd.h
+++ b/gsd/gsd.h
@@ -48,7 +48,10 @@ extern "C"
         GSD_TYPE_FLOAT,
 
         /// 64-bit floating point number.
-        GSD_TYPE_DOUBLE
+        GSD_TYPE_DOUBLE,
+
+        /// 8-bit character.
+        GSD_TYPE_CHARACTER
         };
 
     /// Flag for GSD file open options

From 4ebc9bb4303490308beeeee66cbc92c90f8901a7 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Tue, 16 Jul 2024 16:41:58 -0400
Subject: [PATCH 02/42] Add new type to gsd_sizeof_type

---
 gsd/gsd.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index f20e0f02..2ec94fd0 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -2335,6 +2335,10 @@ size_t gsd_sizeof_type(enum gsd_type type)
         {
         val = sizeof(double);
         }
+    else if (type == GSD_TYPE_CHARACTER)
+        {
+        val = sizeof(char);
+        }
     else
         {
         return 0;

From b11cce5d854cb017d69f1461a5a7818c083cc4af Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Mon, 16 Sep 2024 10:33:28 -0400
Subject: [PATCH 03/42] Add new type to cython files

---
 gsd/fl.pyx     | 142 +++++++++++++++++++++++++++++++++----------------
 gsd/libgsd.pxd |   1 +
 2 files changed, 98 insertions(+), 45 deletions(-)

diff --git a/gsd/fl.pyx b/gsd/fl.pyx
index 03849449..760e1f6f 100644
--- a/gsd/fl.pyx
+++ b/gsd/fl.pyx
@@ -22,6 +22,7 @@ from libc.stdint cimport uint8_t, int8_t, uint16_t, int16_t, uint32_t, int32_t,\
 from libc.errno cimport errno
 cimport gsd.libgsd as libgsd
 cimport numpy
+cimport cython
 
 logger = logging.getLogger('gsd.fl')
 
@@ -145,6 +146,21 @@ cdef void * __get_ptr_float64(data):
     else:
         return <void*>&data_array_float64[0, 0]
 
+cdef void * __get_ptr_char(data):
+    # cdef numpy.ndarray[numpy.str_, ndim=2, mode="c", cast=True] data_array_char
+    cdef size_t address 
+    address = data.__array_interface__["data"][0] 
+    # data_array_char: cython.char[::1] = data.reshape(data.size)
+    # data_array_char = data
+    # cdef char *data_array_char
+    # data_array_char = cython.address(data.data[0])
+    if (data.size == 0):
+        return NULL
+    else:
+        # return <void*>cython.address(data_array_char.data[0])
+        # return <void*>cython.address(data_array_char[0])
+        return <void*>address
+
 
 def open(name, mode, application=None, schema=None, schema_version=None):
     """open(name, mode, application=None, schema=None, schema_version=None)
@@ -557,59 +573,84 @@ cdef class GSDFile:
         if not self.__is_open:
             raise ValueError("File is not open")
 
-        data_array = numpy.ascontiguousarray(data)
-        if data_array is not data:
-            logger.warning('implicit data copy when writing chunk: ' + name)
-        data_array = data_array.view()
-
         cdef uint64_t N
         cdef uint32_t M
 
-        if len(data_array.shape) > 2:
-            raise ValueError("GSD can only write 1 or 2 dimensional arrays: "
-                             + name)
+        cdef libgsd.gsd_type gsd_type
+        cdef void *data_ptr
 
-        if len(data_array.shape) == 1:
-            data_array = data_array.reshape([data_array.shape[0], 1])
+        # Special behavior for handling strings
+        if type(data) is str:
+            bytes_array = numpy.array([data], dtype=numpy.dtype((bytes, len(data))))
+            bytes_view = bytes_array.view(dtype=numpy.int8).reshape((len(data),1))
 
-        N = data_array.shape[0]
-        M = data_array.shape[1]
+            N = len(data)
+            M = 1
 
-        cdef libgsd.gsd_type gsd_type
-        cdef void *data_ptr
-        if data_array.dtype == numpy.uint8:
-            gsd_type = libgsd.GSD_TYPE_UINT8
-            data_ptr = __get_ptr_uint8(data_array)
-        elif data_array.dtype == numpy.uint16:
-            gsd_type = libgsd.GSD_TYPE_UINT16
-            data_ptr = __get_ptr_uint16(data_array)
-        elif data_array.dtype == numpy.uint32:
-            gsd_type = libgsd.GSD_TYPE_UINT32
-            data_ptr = __get_ptr_uint32(data_array)
-        elif data_array.dtype == numpy.uint64:
-            gsd_type = libgsd.GSD_TYPE_UINT64
-            data_ptr = __get_ptr_uint64(data_array)
-        elif data_array.dtype == numpy.int8:
-            gsd_type = libgsd.GSD_TYPE_INT8
-            data_ptr = __get_ptr_int8(data_array)
-        elif data_array.dtype == numpy.int16:
-            gsd_type = libgsd.GSD_TYPE_INT16
-            data_ptr = __get_ptr_int16(data_array)
-        elif data_array.dtype == numpy.int32:
-            gsd_type = libgsd.GSD_TYPE_INT32
-            data_ptr = __get_ptr_int32(data_array)
-        elif data_array.dtype == numpy.int64:
-            gsd_type = libgsd.GSD_TYPE_INT64
-            data_ptr = __get_ptr_int64(data_array)
-        elif data_array.dtype == numpy.float32:
-            gsd_type = libgsd.GSD_TYPE_FLOAT
-            data_ptr = __get_ptr_float32(data_array)
-        elif data_array.dtype == numpy.float64:
-            gsd_type = libgsd.GSD_TYPE_DOUBLE
-            data_ptr = __get_ptr_float64(data_array)
+            gsd_type = libgsd.GSD_TYPE_CHARACTER
+            data_ptr = __get_ptr_int8(bytes_view)
+
+        # Non-string behavior
         else:
-            raise ValueError("invalid type for chunk: " + name)
+            data_array = numpy.ascontiguousarray(data)
+
+            if data_array is not data:
+                logger.warning('implicit data copy when writing chunk: ' + name)
+            data_array = data_array.view()
+
+
+
+            if len(data_array.shape) > 2:
+                raise ValueError("GSD can only write 1 or 2 dimensional arrays: "
+                                + name)
+
+            if len(data_array.shape) == 1:
+                data_array = data_array.reshape([data_array.shape[0], 1])
 
+            N = data_array.shape[0]
+            M = data_array.shape[1]
+
+            if data_array.dtype == numpy.uint8:
+                gsd_type = libgsd.GSD_TYPE_UINT8
+                data_ptr = __get_ptr_uint8(data_array)
+            elif data_array.dtype == numpy.uint16:
+                gsd_type = libgsd.GSD_TYPE_UINT16
+                data_ptr = __get_ptr_uint16(data_array)
+            elif data_array.dtype == numpy.uint32:
+                gsd_type = libgsd.GSD_TYPE_UINT32
+                data_ptr = __get_ptr_uint32(data_array)
+            elif data_array.dtype == numpy.uint64:
+                gsd_type = libgsd.GSD_TYPE_UINT64
+                data_ptr = __get_ptr_uint64(data_array)
+            elif data_array.dtype == numpy.int8:
+                gsd_type = libgsd.GSD_TYPE_INT8
+                data_ptr = __get_ptr_int8(data_array)
+            elif data_array.dtype == numpy.int16:
+                gsd_type = libgsd.GSD_TYPE_INT16
+                data_ptr = __get_ptr_int16(data_array)
+            elif data_array.dtype == numpy.int32:
+                gsd_type = libgsd.GSD_TYPE_INT32
+                data_ptr = __get_ptr_int32(data_array)
+            elif data_array.dtype == numpy.int64:
+                gsd_type = libgsd.GSD_TYPE_INT64
+                data_ptr = __get_ptr_int64(data_array)
+            elif data_array.dtype == numpy.float32:
+                gsd_type = libgsd.GSD_TYPE_FLOAT
+                data_ptr = __get_ptr_float32(data_array)
+            elif data_array.dtype == numpy.float64:
+                gsd_type = libgsd.GSD_TYPE_DOUBLE
+                data_ptr = __get_ptr_float64(data_array)
+            # elif data_array.dtype.type is numpy.str_:
+            #     if N != 0:
+            #         M = int(data_array.dtype.itemsize/4)
+            #     print(f"Wrote: {N=}, {M=}, {data_array=}")
+            #     gsd_type = libgsd.GSD_TYPE_CHARACTER
+            #     data_ptr = __get_ptr_char(data_array)
+            else:
+                raise ValueError("invalid type for chunk: " + name)
+
+        # Once we have the data pointer, the behavior should be identical
+        # for all data types
         logger.debug('write chunk: ' + self.name + ' - ' + name)
 
         cdef char * c_name
@@ -787,6 +828,9 @@ cdef class GSDFile:
         elif gsd_type == libgsd.GSD_TYPE_DOUBLE:
             data_array = numpy.empty(dtype=numpy.float64,
                                      shape=[index_entry.N, index_entry.M])
+        elif gsd_type == libgsd.GSD_TYPE_CHARACTER:
+            data_array = numpy.empty(dtype=numpy.int8,
+                                     shape=[index_entry.M, index_entry.N])
         else:
             raise ValueError("invalid type for chunk: " + name)
 
@@ -815,6 +859,8 @@ cdef class GSDFile:
                 data_ptr = __get_ptr_float32(data_array)
             elif gsd_type == libgsd.GSD_TYPE_DOUBLE:
                 data_ptr = __get_ptr_float64(data_array)
+            elif gsd_type == libgsd.GSD_TYPE_CHARACTER:
+                data_ptr = __get_ptr_int8(data_array)
             else:
                 raise ValueError("invalid type for chunk: " + name)
 
@@ -826,6 +872,12 @@ cdef class GSDFile:
             __raise_on_error(retval, self.name)
 
         if index_entry.M == 1:
+            if gsd_type == libgsd.GSD_TYPE_CHARACTER:
+                data_array = data_array.flatten()
+                bytes_array = data_array.view(dtype=numpy.dtype((bytes, data_array.shape[0])))
+                print(bytes_array[0].decode("UTF-8"))
+                return bytes_array[0].decode("UTF-8")
+            
             return data_array.reshape([index_entry.N])
         else:
             return data_array
diff --git a/gsd/libgsd.pxd b/gsd/libgsd.pxd
index e8bb79d8..1d42d078 100644
--- a/gsd/libgsd.pxd
+++ b/gsd/libgsd.pxd
@@ -16,6 +16,7 @@ cdef extern from "gsd.h" nogil:
         GSD_TYPE_INT64
         GSD_TYPE_FLOAT
         GSD_TYPE_DOUBLE
+        GSD_TYPE_CHARACTER
 
     cdef enum gsd_open_flag:
         GSD_OPEN_READWRITE=1

From 06698117f1d4ba191c73bb855e9f2cd5c2ce7362 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Mon, 16 Sep 2024 10:33:58 -0400
Subject: [PATCH 04/42] Add new type to pygsd

---
 gsd/pygsd.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index f21c1177..00ef0145 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -63,6 +63,7 @@
     8: numpy.dtype('int64'),
     9: numpy.dtype('float32'),
     10: numpy.dtype('float64'),
+    11: numpy.dtype('str_'),
 }
 
 
@@ -352,7 +353,7 @@ def read_chunk(self, frame, name):
 
         if len(data_raw) != size:
             raise OSError
-
+        # TODO: if gsd type is character, decode it here
         data_npy = numpy.frombuffer(data_raw, dtype=gsd_type_mapping[chunk.type])
 
         if chunk.M == 1:

From 6b2fdf719a47120ba722fe44b4b504029c965bc7 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Mon, 16 Sep 2024 10:34:23 -0400
Subject: [PATCH 05/42] Create string test for fl

---
 gsd/test/test_fl.py | 37 +++++++++++++++++++++++--------------
 1 file changed, 23 insertions(+), 14 deletions(-)

diff --git a/gsd/test/test_fl.py b/gsd/test/test_fl.py
index 57a18484..9b1b4679 100644
--- a/gsd/test/test_fl.py
+++ b/gsd/test/test_fl.py
@@ -43,12 +43,16 @@ def test_create(tmp_path, open_mode):
         numpy.int64,
         numpy.float32,
         numpy.float64,
+        numpy.str_,
     ],
 )
 def test_dtype(tmp_path, typ):
     """Test all supported data types."""
-    data1d = numpy.array([1, 2, 3, 4, 5, 127], dtype=typ)
-    data2d = numpy.array([[10, 20], [30, 40], [50, 80]], dtype=typ)
+    if typ is numpy.str_:
+        data1d = "test"
+    else:
+        data1d = numpy.array([950, 200, 3, 4, 5, 6, 7], dtype=typ)
+    data2d = numpy.array([[10, 20], [30, 40], [50, 80]], dtype=typ) # <-- don't use this with str
     data_zero = numpy.array([], dtype=typ)
 
     gsd.fl.open(
@@ -67,8 +71,9 @@ def test_dtype(tmp_path, typ):
         schema_version=[1, 2],
     ) as f:
         f.write_chunk(name='data1d', data=data1d)
-        f.write_chunk(name='data2d', data=data2d)
-        f.write_chunk(name='data_zero', data=data_zero)
+        if typ is not numpy.str_:
+            f.write_chunk(name='data2d', data=data2d)
+            f.write_chunk(name='data_zero', data=data_zero)
         f.end_frame()
 
     with gsd.fl.open(
@@ -79,24 +84,28 @@ def test_dtype(tmp_path, typ):
         schema_version=[1, 2],
     ) as f:
         read_data1d = f.read_chunk(frame=0, name='data1d')
-        read_data2d = f.read_chunk(frame=0, name='data2d')
-        read_data_zero = f.read_chunk(frame=0, name='data_zero')
-
-        assert data1d.dtype == read_data1d.dtype
+        if typ is not numpy.str_:
+            read_data2d = f.read_chunk(frame=0, name='data2d')
+            read_data_zero = f.read_chunk(frame=0, name='data_zero')
+
+        if typ != numpy.str_:
+            assert data1d.dtype.type == read_data1d.dtype.type
+            assert data2d.dtype.type == read_data2d.dtype.type
+            numpy.testing.assert_array_equal(data2d, read_data2d)
+            assert data_zero.dtype.type == read_data_zero.dtype.type
+            assert data_zero.shape == (0,)
+        
         numpy.testing.assert_array_equal(data1d, read_data1d)
-        assert data2d.dtype == read_data2d.dtype
-        numpy.testing.assert_array_equal(data2d, read_data2d)
-        assert data_zero.dtype == read_data_zero.dtype
-        assert data_zero.shape == (0,)
+
 
     # test again with pygsd
     with gsd.pygsd.GSDFile(file=open(str(tmp_path / 'test_dtype.gsd'), mode='rb')) as f:
         read_data1d = f.read_chunk(frame=0, name='data1d')
         read_data2d = f.read_chunk(frame=0, name='data2d')
 
-        assert data1d.dtype == read_data1d.dtype
+        assert data1d.dtype.type == read_data1d.dtype.type
         numpy.testing.assert_array_equal(data1d, read_data1d)
-        assert data2d.dtype == read_data2d.dtype
+        assert data2d.dtype.type == read_data2d.dtype.type
         numpy.testing.assert_array_equal(data2d, read_data2d)
 
 

From 4c44b4fe93235999b97143e7ced4fda35ff516f1 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Mon, 16 Sep 2024 12:25:06 -0400
Subject: [PATCH 06/42] Split dtype tests into string and nonstring

---
 gsd/test/test_fl.py | 77 +++++++++++++++++++++++++++++++--------------
 1 file changed, 54 insertions(+), 23 deletions(-)

diff --git a/gsd/test/test_fl.py b/gsd/test/test_fl.py
index 9b1b4679..69fdd0c1 100644
--- a/gsd/test/test_fl.py
+++ b/gsd/test/test_fl.py
@@ -43,16 +43,12 @@ def test_create(tmp_path, open_mode):
         numpy.int64,
         numpy.float32,
         numpy.float64,
-        numpy.str_,
     ],
 )
-def test_dtype(tmp_path, typ):
-    """Test all supported data types."""
-    if typ is numpy.str_:
-        data1d = "test"
-    else:
-        data1d = numpy.array([950, 200, 3, 4, 5, 6, 7], dtype=typ)
-    data2d = numpy.array([[10, 20], [30, 40], [50, 80]], dtype=typ) # <-- don't use this with str
+def test_nonstring_dtypes(tmp_path, typ):
+    """Test all supported data types except for strings."""
+    data1d = numpy.array([1, 2, 3, 4, 5, 127], dtype=typ)
+    data2d = numpy.array([[10, 20], [30, 40], [50, 80]], dtype=typ)
     data_zero = numpy.array([], dtype=typ)
 
     gsd.fl.open(
@@ -71,9 +67,8 @@ def test_dtype(tmp_path, typ):
         schema_version=[1, 2],
     ) as f:
         f.write_chunk(name='data1d', data=data1d)
-        if typ is not numpy.str_:
-            f.write_chunk(name='data2d', data=data2d)
-            f.write_chunk(name='data_zero', data=data_zero)
+        f.write_chunk(name='data2d', data=data2d)
+        f.write_chunk(name='data_zero', data=data_zero)
         f.end_frame()
 
     with gsd.fl.open(
@@ -84,19 +79,15 @@ def test_dtype(tmp_path, typ):
         schema_version=[1, 2],
     ) as f:
         read_data1d = f.read_chunk(frame=0, name='data1d')
-        if typ is not numpy.str_:
-            read_data2d = f.read_chunk(frame=0, name='data2d')
-            read_data_zero = f.read_chunk(frame=0, name='data_zero')
-
-        if typ != numpy.str_:
-            assert data1d.dtype.type == read_data1d.dtype.type
-            assert data2d.dtype.type == read_data2d.dtype.type
-            numpy.testing.assert_array_equal(data2d, read_data2d)
-            assert data_zero.dtype.type == read_data_zero.dtype.type
-            assert data_zero.shape == (0,)
-        
-        numpy.testing.assert_array_equal(data1d, read_data1d)
+        read_data2d = f.read_chunk(frame=0, name='data2d')
+        read_data_zero = f.read_chunk(frame=0, name='data_zero')
 
+        assert data1d.dtype.type == read_data1d.dtype.type
+        numpy.testing.assert_array_equal(data1d, read_data1d)
+        assert data2d.dtype.type == read_data2d.dtype.type
+        numpy.testing.assert_array_equal(data2d, read_data2d)
+        assert data_zero.dtype.type == read_data_zero.dtype.type
+        assert data_zero.shape == (0,)
 
     # test again with pygsd
     with gsd.pygsd.GSDFile(file=open(str(tmp_path / 'test_dtype.gsd'), mode='rb')) as f:
@@ -108,6 +99,46 @@ def test_dtype(tmp_path, typ):
         assert data2d.dtype.type == read_data2d.dtype.type
         numpy.testing.assert_array_equal(data2d, read_data2d)
 
+def test_string_dtype(tmp_path):
+    """Test string datatype.
+    
+    Note that the string datatype does not support 0-D or 2-D data."""
+    data1d = "test"
+
+    gsd.fl.open(
+        mode='x',
+        name=tmp_path / 'test_dtype.gsd',
+        application='test_dtype',
+        schema='none',
+        schema_version=[1, 2],
+    )
+
+    with gsd.fl.open(
+        name=tmp_path / 'test_dtype.gsd',
+        mode='w',
+        application='test_dtype',
+        schema='none',
+        schema_version=[1, 2],
+    ) as f:
+        f.write_chunk(name='data1d', data=data1d)
+        f.end_frame()
+
+    with gsd.fl.open(
+        name=tmp_path / 'test_dtype.gsd',
+        mode='r',
+        application='test_dtype',
+        schema='none',
+        schema_version=[1, 2],
+    ) as f:
+        read_data1d = f.read_chunk(frame=0, name='data1d')
+
+        numpy.testing.assert_string_equal(data1d, read_data1d)
+
+    # test again with pygsd
+    with gsd.pygsd.GSDFile(file=open(str(tmp_path / 'test_dtype.gsd'), mode='rb')) as f:
+        read_data1d = f.read_chunk(frame=0, name='data1d')
+
+        numpy.testing.assert_string_equal(data1d, read_data1d)
 
 def test_metadata(tmp_path, open_mode):
     """Test file metadata."""

From 45c0e961d818987ba941554f05b6dfa923d33bd7 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Mon, 16 Sep 2024 12:25:25 -0400
Subject: [PATCH 07/42] Add string decoding to pygsd

---
 gsd/pygsd.py | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index 9ba5e7ea..ea1cf771 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -63,7 +63,7 @@
     8: numpy.dtype('int64'),
     9: numpy.dtype('float32'),
     10: numpy.dtype('float64'),
-    11: numpy.dtype('str_'),
+    11: numpy.dtype('int8'),    # used for strings
 }
 
 
@@ -353,8 +353,12 @@ def read_chunk(self, frame, name):
 
         if len(data_raw) != size:
             raise OSError
-        # TODO: if gsd type is character, decode it here
-        data_npy = numpy.frombuffer(data_raw, dtype=gsd_type_mapping[chunk.type])
+        
+        # If gsd type is character, decode it here
+        if chunk.type == 11:
+            data_npy = data_raw.decode('utf-8')
+        else:
+            data_npy = numpy.frombuffer(data_raw, dtype=gsd_type_mapping[chunk.type])
 
         if chunk.M == 1:
             return data_npy

From 21a63eff32261c4aa7d712c1e90a957136c245bf Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Mon, 16 Sep 2024 12:26:37 -0400
Subject: [PATCH 08/42] Add note about strings to the file layer docs

---
 doc/file-layer.rst | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/doc/file-layer.rst b/doc/file-layer.rst
index 78431216..3f416a2b 100644
--- a/doc/file-layer.rst
+++ b/doc/file-layer.rst
@@ -235,3 +235,6 @@ Data block
 A data block stores raw data bytes on the disk. For a given index entry
 ``entry``, the data starts at location ``entry.location`` and is the next
 ``entry.N * entry.M * gsd_sizeof_type(entry.type)`` bytes.
+
+The type fields are self-explanatory except for UTF-8 strings, for which
+null-termination is allowed but not required.
\ No newline at end of file

From 58c16a4c6a485542b642c85e7f541c60de21b080 Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Mon, 16 Sep 2024 17:25:36 +0000
Subject: [PATCH 09/42] [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci
---
 doc/file-layer.rst  | 2 +-
 gsd/fl.pyx          | 6 +++---
 gsd/pygsd.py        | 4 ++--
 gsd/test/test_fl.py | 6 ++++--
 4 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/doc/file-layer.rst b/doc/file-layer.rst
index 3f416a2b..96c1b50d 100644
--- a/doc/file-layer.rst
+++ b/doc/file-layer.rst
@@ -237,4 +237,4 @@ A data block stores raw data bytes on the disk. For a given index entry
 ``entry.N * entry.M * gsd_sizeof_type(entry.type)`` bytes.
 
 The type fields are self-explanatory except for UTF-8 strings, for which
-null-termination is allowed but not required.
\ No newline at end of file
+null-termination is allowed but not required.
diff --git a/gsd/fl.pyx b/gsd/fl.pyx
index 760e1f6f..cb9bbc16 100644
--- a/gsd/fl.pyx
+++ b/gsd/fl.pyx
@@ -148,8 +148,8 @@ cdef void * __get_ptr_float64(data):
 
 cdef void * __get_ptr_char(data):
     # cdef numpy.ndarray[numpy.str_, ndim=2, mode="c", cast=True] data_array_char
-    cdef size_t address 
-    address = data.__array_interface__["data"][0] 
+    cdef size_t address
+    address = data.__array_interface__["data"][0]
     # data_array_char: cython.char[::1] = data.reshape(data.size)
     # data_array_char = data
     # cdef char *data_array_char
@@ -877,7 +877,7 @@ cdef class GSDFile:
                 bytes_array = data_array.view(dtype=numpy.dtype((bytes, data_array.shape[0])))
                 print(bytes_array[0].decode("UTF-8"))
                 return bytes_array[0].decode("UTF-8")
-            
+
             return data_array.reshape([index_entry.N])
         else:
             return data_array
diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index ea1cf771..00a750ae 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -63,7 +63,7 @@
     8: numpy.dtype('int64'),
     9: numpy.dtype('float32'),
     10: numpy.dtype('float64'),
-    11: numpy.dtype('int8'),    # used for strings
+    11: numpy.dtype('int8'),  # used for strings
 }
 
 
@@ -353,7 +353,7 @@ def read_chunk(self, frame, name):
 
         if len(data_raw) != size:
             raise OSError
-        
+
         # If gsd type is character, decode it here
         if chunk.type == 11:
             data_npy = data_raw.decode('utf-8')
diff --git a/gsd/test/test_fl.py b/gsd/test/test_fl.py
index 69fdd0c1..844db625 100644
--- a/gsd/test/test_fl.py
+++ b/gsd/test/test_fl.py
@@ -99,11 +99,12 @@ def test_nonstring_dtypes(tmp_path, typ):
         assert data2d.dtype.type == read_data2d.dtype.type
         numpy.testing.assert_array_equal(data2d, read_data2d)
 
+
 def test_string_dtype(tmp_path):
     """Test string datatype.
-    
+
     Note that the string datatype does not support 0-D or 2-D data."""
-    data1d = "test"
+    data1d = 'test'
 
     gsd.fl.open(
         mode='x',
@@ -140,6 +141,7 @@ def test_string_dtype(tmp_path):
 
         numpy.testing.assert_string_equal(data1d, read_data1d)
 
+
 def test_metadata(tmp_path, open_mode):
     """Test file metadata."""
     data = numpy.array([1, 2, 3, 4, 5, 10012], dtype=numpy.int64)

From e12a9a240b1552a8bf28cecbd3c66dc7ce75973b Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Wed, 18 Sep 2024 09:34:52 -0400
Subject: [PATCH 10/42] Revise docstring for pre-commit

---
 gsd/test/test_fl.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/gsd/test/test_fl.py b/gsd/test/test_fl.py
index 844db625..5ec32a30 100644
--- a/gsd/test/test_fl.py
+++ b/gsd/test/test_fl.py
@@ -102,9 +102,10 @@ def test_nonstring_dtypes(tmp_path, typ):
 
 def test_string_dtype(tmp_path):
     """Test string datatype.
-
-    Note that the string datatype does not support 0-D or 2-D data."""
-    data1d = 'test'
+    
+    Note that the string datatype does not support 0-D or 2-D data.
+    """
+    data1d = "test"
 
     gsd.fl.open(
         mode='x',

From d1ad023f64f9c557cae649a9fdc4a5807590cc68 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Wed, 18 Sep 2024 10:23:36 -0400
Subject: [PATCH 11/42] Replace type mapping dict with enum using class method

---
 gsd/pygsd.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index 00a750ae..235539ea 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -29,6 +29,7 @@
 
 """
 
+from enum import Enum
 import logging
 import struct
 import sys

From a7aa0c2b356e066bcaf953266feef820f50f1b81 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Wed, 18 Sep 2024 10:24:13 -0400
Subject: [PATCH 12/42] Switch enum to use functional approach

---
 gsd/pygsd.py | 29 ++++++++++++++++-------------
 1 file changed, 16 insertions(+), 13 deletions(-)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index 235539ea..f2546aba 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -53,19 +53,22 @@
 gsd_index_entry = namedtuple('gsd_index_entry', 'frame N location M id type flags')
 gsd_index_entry_struct = struct.Struct('QQqIHBB')
 
-gsd_type_mapping = {
-    1: numpy.dtype('uint8'),
-    2: numpy.dtype('uint16'),
-    3: numpy.dtype('uint32'),
-    4: numpy.dtype('uint64'),
-    5: numpy.dtype('int8'),
-    6: numpy.dtype('int16'),
-    7: numpy.dtype('int32'),
-    8: numpy.dtype('int64'),
-    9: numpy.dtype('float32'),
-    10: numpy.dtype('float64'),
-    11: numpy.dtype('int8'),  # used for strings
-}
+gsd_type_mapping = Enum(
+    "gsd_type",
+    [
+        ["uint8", numpy.dtype('uint8')],
+        ["uint16", numpy.dtype('uint16')],
+        ["uint32", numpy.dtype('uint32')],
+        ["uint64", numpy.dtype('uint64')],
+        ["int8", numpy.dtype('int8')],
+        ["int16", numpy.dtype('int16')],
+        ["int32", numpy.dtype('int32')],
+        ["int64", numpy.dtype('int64')],
+        ["float32", numpy.dtype('float32')],
+        ["float64", numpy.dtype('float64')],
+        ["str", numpy.dtype('int8')],    # used for strings
+    ]
+)
 
 
 class GSDFile:

From 4c4099cb679f4a56f44e67eefe16b157af533cf8 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 08:54:01 -0400
Subject: [PATCH 13/42] Replace Enum with map of tuples

---
 gsd/pygsd.py | 37 +++++++++++++++++--------------------
 1 file changed, 17 insertions(+), 20 deletions(-)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index f2546aba..c8f54d8e 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -53,22 +53,19 @@
 gsd_index_entry = namedtuple('gsd_index_entry', 'frame N location M id type flags')
 gsd_index_entry_struct = struct.Struct('QQqIHBB')
 
-gsd_type_mapping = Enum(
-    "gsd_type",
-    [
-        ["uint8", numpy.dtype('uint8')],
-        ["uint16", numpy.dtype('uint16')],
-        ["uint32", numpy.dtype('uint32')],
-        ["uint64", numpy.dtype('uint64')],
-        ["int8", numpy.dtype('int8')],
-        ["int16", numpy.dtype('int16')],
-        ["int32", numpy.dtype('int32')],
-        ["int64", numpy.dtype('int64')],
-        ["float32", numpy.dtype('float32')],
-        ["float64", numpy.dtype('float64')],
-        ["str", numpy.dtype('int8')],    # used for strings
-    ]
-)
+gsd_type_mapping = {
+    1: ("uint8", numpy.dtype('uint8')),
+    2: ("uint16", numpy.dtype('uint16')),
+    3: ("uint32", numpy.dtype('uint32')),
+    4: ("uint64", numpy.dtype('uint64')),
+    5: ("int8", numpy.dtype('int8')),
+    6: ("int16", numpy.dtype('int16')),
+    7: ("int32", numpy.dtype('int32')),
+    8: ("int64", numpy.dtype('int64')),
+    9: ("float32", numpy.dtype('float32')),
+    10: ("float64", numpy.dtype('float64')),
+    11: ("str", numpy.dtype('int8')),
+}
 
 
 class GSDFile:
@@ -338,7 +335,7 @@ def read_chunk(self, frame, name):
             'read chunk: ' + str(self.__file) + ' - ' + str(frame) + ' - ' + name
         )
 
-        size = chunk.N * chunk.M * gsd_type_mapping[chunk.type].itemsize
+        size = chunk.N * chunk.M * gsd_type_mapping[chunk.type][1].itemsize
         if chunk.location == 0:
             raise RuntimeError(
                 'Corrupt chunk: '
@@ -350,7 +347,7 @@ def read_chunk(self, frame, name):
             )
 
         if size == 0:
-            return numpy.array([], dtype=gsd_type_mapping[chunk.type])
+            return numpy.array([], dtype=gsd_type_mapping[chunk.type][1])
 
         self.__file.seek(chunk.location, 0)
         data_raw = self.__file.read(size)
@@ -359,10 +356,10 @@ def read_chunk(self, frame, name):
             raise OSError
 
         # If gsd type is character, decode it here
-        if chunk.type == 11:
+        if gsd_type_mapping[chunk.type][0] == "str":
             data_npy = data_raw.decode('utf-8')
         else:
-            data_npy = numpy.frombuffer(data_raw, dtype=gsd_type_mapping[chunk.type])
+            data_npy = numpy.frombuffer(data_raw, dtype=gsd_type_mapping[chunk.type][1])
 
         if chunk.M == 1:
             return data_npy

From 45a6ce86d4376c877b3cf1622b63ad27c7c86755 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 09:09:40 -0400
Subject: [PATCH 14/42] Fix formatting with pre-commit

---
 gsd/test/test_fl.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/gsd/test/test_fl.py b/gsd/test/test_fl.py
index 5ec32a30..7ea6d069 100644
--- a/gsd/test/test_fl.py
+++ b/gsd/test/test_fl.py
@@ -102,10 +102,10 @@ def test_nonstring_dtypes(tmp_path, typ):
 
 def test_string_dtype(tmp_path):
     """Test string datatype.
-    
+
     Note that the string datatype does not support 0-D or 2-D data.
     """
-    data1d = "test"
+    data1d = 'test'
 
     gsd.fl.open(
         mode='x',

From 0b6ff900dc7d0ce325a79e5c4513147a7dd68830 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 09:09:56 -0400
Subject: [PATCH 15/42] Fix formatting with pre-commit

---
 gsd/pygsd.py | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index c8f54d8e..e6bcaae1 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -54,17 +54,17 @@
 gsd_index_entry_struct = struct.Struct('QQqIHBB')
 
 gsd_type_mapping = {
-    1: ("uint8", numpy.dtype('uint8')),
-    2: ("uint16", numpy.dtype('uint16')),
-    3: ("uint32", numpy.dtype('uint32')),
-    4: ("uint64", numpy.dtype('uint64')),
-    5: ("int8", numpy.dtype('int8')),
-    6: ("int16", numpy.dtype('int16')),
-    7: ("int32", numpy.dtype('int32')),
-    8: ("int64", numpy.dtype('int64')),
-    9: ("float32", numpy.dtype('float32')),
-    10: ("float64", numpy.dtype('float64')),
-    11: ("str", numpy.dtype('int8')),
+    1: ('uint8', numpy.dtype('uint8')),
+    2: ('uint16', numpy.dtype('uint16')),
+    3: ('uint32', numpy.dtype('uint32')),
+    4: ('uint64', numpy.dtype('uint64')),
+    5: ('int8', numpy.dtype('int8')),
+    6: ('int16', numpy.dtype('int16')),
+    7: ('int32', numpy.dtype('int32')),
+    8: ('int64', numpy.dtype('int64')),
+    9: ('float32', numpy.dtype('float32')),
+    10: ('float64', numpy.dtype('float64')),
+    11: ('str', numpy.dtype('int8')),
 }
 
 

From 385d6a1549868c0521003d1a325cb04ad6ef444f Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 09:10:22 -0400
Subject: [PATCH 16/42] Remove unused import

---
 gsd/pygsd.py | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/gsd/pygsd.py b/gsd/pygsd.py
index e6bcaae1..9e1adfec 100644
--- a/gsd/pygsd.py
+++ b/gsd/pygsd.py
@@ -29,7 +29,6 @@
 
 """
 
-from enum import Enum
 import logging
 import struct
 import sys
@@ -356,7 +355,7 @@ def read_chunk(self, frame, name):
             raise OSError
 
         # If gsd type is character, decode it here
-        if gsd_type_mapping[chunk.type][0] == "str":
+        if gsd_type_mapping[chunk.type][0] == 'str':
             data_npy = data_raw.decode('utf-8')
         else:
             data_npy = numpy.frombuffer(data_raw, dtype=gsd_type_mapping[chunk.type][1])

From 6f564687d0e308746a990efa2a237d349df68fc9 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 09:54:24 -0400
Subject: [PATCH 17/42] Silently upgrade minor file version on write

---
 gsd/gsd.c | 30 ++++++++++++++++++++++++++----
 1 file changed, 26 insertions(+), 4 deletions(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index 5a4bbac0..ef2db77e 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -86,7 +86,12 @@ enum
 /// Current GSD file specification
 enum
     {
-    GSD_CURRENT_FILE_VERSION = 2
+    GSD_CURRENT_FILE_VERSION_MAJOR = 2
+    };
+
+enum
+    {
+    GSD_CURRENT_FILE_VERSION_MINOR = 1
     };
 
 // define windows wrapper functions
@@ -1384,7 +1389,8 @@ gsd_initialize_file(int fd, const char* application, const char* schema, uint32_
     gsd_util_zero_memory(&header, sizeof(header));
 
     header.magic = GSD_MAGIC_ID;
-    header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION, 0);
+    header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, 
+                                          GSD_CURRENT_FILE_VERSION_MINOR);
     strncpy(header.application, application, sizeof(header.application) - 1);
     header.application[sizeof(header.application) - 1] = 0;
     strncpy(header.schema, schema, sizeof(header.schema) - 1);
@@ -1607,6 +1613,21 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
     handle->maximum_write_buffer_size = GSD_DEFAULT_MAXIMUM_WRITE_BUFFER_SIZE;
     handle->index_entries_to_buffer = GSD_DEFAULT_INDEX_ENTRIES_TO_BUFFER;
 
+    // When opening a file in a writeable mode, if the file's major version
+    // is identical to the current major version, silently upgrade the minor version
+    if((handle->open_flags == GSD_OPEN_READWRITE || handle-> open_flags == GSD_OPEN_APPEND) && 
+       (handle->header.gsd_version != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR)) &&
+       (handle->header.gsd_version >> 16 == GSD_CURRENT_FILE_VERSION_MAJOR)) {
+        
+        handle->header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);
+        size_t bytes_written = gsd_io_pwrite_retry(handle->fd, &(handle->header), sizeof(struct gsd_header), 0);
+        
+        if (bytes_written != sizeof(struct gsd_header))
+            {
+            return GSD_ERROR_IO;
+            }
+       }
+
     return GSD_SUCCESS;
     }
 
@@ -2558,8 +2579,9 @@ int gsd_upgrade(struct gsd_handle* handle)
                 }
             }
 
-        // label the file as a v2.0 file
-        handle->header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION, 0);
+        // label the file as a v2.1 file
+        handle->header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, 
+                                                      GSD_CURRENT_FILE_VERSION_MINOR);
 
         // write the new header out
         ssize_t bytes_written

From 75595dec13bceff09790b03d516981ef33cba1da Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 09:56:07 -0400
Subject: [PATCH 18/42] Fix formatting with pre-commit

---
 gsd/gsd.c | 27 +++++++++++++++------------
 1 file changed, 15 insertions(+), 12 deletions(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index ef2db77e..bd9de68b 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1389,8 +1389,8 @@ gsd_initialize_file(int fd, const char* application, const char* schema, uint32_
     gsd_util_zero_memory(&header, sizeof(header));
 
     header.magic = GSD_MAGIC_ID;
-    header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, 
-                                          GSD_CURRENT_FILE_VERSION_MINOR);
+    header.gsd_version
+        = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);
     strncpy(header.application, application, sizeof(header.application) - 1);
     header.application[sizeof(header.application) - 1] = 0;
     strncpy(header.schema, schema, sizeof(header.schema) - 1);
@@ -1615,18 +1615,21 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
 
     // When opening a file in a writeable mode, if the file's major version
     // is identical to the current major version, silently upgrade the minor version
-    if((handle->open_flags == GSD_OPEN_READWRITE || handle-> open_flags == GSD_OPEN_APPEND) && 
-       (handle->header.gsd_version != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR)) &&
-       (handle->header.gsd_version >> 16 == GSD_CURRENT_FILE_VERSION_MAJOR)) {
-        
-        handle->header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);
-        size_t bytes_written = gsd_io_pwrite_retry(handle->fd, &(handle->header), sizeof(struct gsd_header), 0);
-        
+    if ((handle->open_flags == GSD_OPEN_READWRITE || handle->open_flags == GSD_OPEN_APPEND)
+        && (handle->header.gsd_version
+            != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
+        && (handle->header.gsd_version >> 16 == GSD_CURRENT_FILE_VERSION_MAJOR))
+        {
+        handle->header.gsd_version
+            = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);
+        size_t bytes_written
+            = gsd_io_pwrite_retry(handle->fd, &(handle->header), sizeof(struct gsd_header), 0);
+
         if (bytes_written != sizeof(struct gsd_header))
             {
             return GSD_ERROR_IO;
             }
-       }
+        }
 
     return GSD_SUCCESS;
     }
@@ -2580,8 +2583,8 @@ int gsd_upgrade(struct gsd_handle* handle)
             }
 
         // label the file as a v2.1 file
-        handle->header.gsd_version = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, 
-                                                      GSD_CURRENT_FILE_VERSION_MINOR);
+        handle->header.gsd_version
+            = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);
 
         // write the new header out
         ssize_t bytes_written

From 26cdc20ece9339ae66a31ed2d7041507ea0b2f2e Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 09:56:27 -0400
Subject: [PATCH 19/42] Abstract hard-coded current version into separate
 variable

---
 gsd/test/test_fl.py | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/gsd/test/test_fl.py b/gsd/test/test_fl.py
index 7ea6d069..fe92c853 100644
--- a/gsd/test/test_fl.py
+++ b/gsd/test/test_fl.py
@@ -17,6 +17,7 @@
 import gsd.pygsd
 
 test_path = pathlib.Path(os.path.realpath(__file__)).parent
+current_gsd_version = (2, 1)
 
 
 def test_create(tmp_path, open_mode):
@@ -172,7 +173,7 @@ def test_metadata(tmp_path, open_mode):
         assert f.schema == 'none'
         assert f.schema_version == (1, 2)
         assert f.nframes == 150
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
     # test again with pygsd
     with gsd.pygsd.GSDFile(
@@ -184,7 +185,7 @@ def test_metadata(tmp_path, open_mode):
         assert f.schema == 'none'
         assert f.schema_version == (1, 2)
         assert f.nframes == 150
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
 
 def test_append(tmp_path, open_mode):
@@ -815,14 +816,14 @@ def check_v1_file_read(f):
         schema='none',
         schema_version=[1, 2],
     ) as f:
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
         check_v1_file_read(f)
 
     with gsd.pygsd.GSDFile(
         file=open(str(tmp_path / 'test_gsd_v1.gsd'), mode='rb')
     ) as f:
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
         check_v1_file_read(f)
 
@@ -955,7 +956,7 @@ def check_v1_file_read(f):
 
         f.upgrade()
 
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
         for value in values:
             if isinstance(value, int):
@@ -975,7 +976,7 @@ def check_v1_file_read(f):
         schema='none',
         schema_version=[1, 2],
     ) as f:
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
         check_v1_file_read(f)
 
@@ -983,7 +984,7 @@ def check_v1_file_read(f):
     with gsd.pygsd.GSDFile(
         file=open(str(tmp_path / 'test_gsd_v1.gsd'), mode='rb')
     ) as f:
-        assert f.gsd_version == (2, 0)
+        assert f.gsd_version == current_gsd_version
 
         check_v1_file_read(f)
 

From fa57eaedbccb4a1c039d7c3bead726fef454195a Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 10:00:40 -0400
Subject: [PATCH 20/42] Update changelog

---
 CHANGELOG.rst | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index 4d0e30b4..70523a6e 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -10,6 +10,14 @@ Change Log
 3.x
 ---
 
+3.4.0 (not yet released)
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+*Added:*
+
+* New chunk type for string data - valid in file layer versions 2.1 and later
+  (`#391 <https://github.com/glotzerlab/gsd/pull/391>`__).
+
 3.3.2 (2024-09-06)
 ^^^^^^^^^^^^^^^^^^
 

From 138ce40571918cab2cd8e9170d3a06e2d6215bd6 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 10:00:47 -0400
Subject: [PATCH 21/42] Update credits

---
 doc/credits.rst | 1 +
 1 file changed, 1 insertion(+)

diff --git a/doc/credits.rst b/doc/credits.rst
index 2d7c5237..298b157f 100644
--- a/doc/credits.rst
+++ b/doc/credits.rst
@@ -18,3 +18,4 @@ The following people contributed to GSD.
 * Alexander Stukowski, OVITO GmbH
 * Charlotte Shiqi Zhao, University of Michigan
 * Tim Moore, University of Michigan
+* Joseph Burkhart, University of Michigan
\ No newline at end of file

From 28551fc35c9f8b2c60f44e33f6eabf8185ce9259 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 10:01:32 -0400
Subject: [PATCH 22/42] Fix formatting with pre-commit

---
 doc/credits.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/credits.rst b/doc/credits.rst
index 298b157f..18f9f19a 100644
--- a/doc/credits.rst
+++ b/doc/credits.rst
@@ -18,4 +18,4 @@ The following people contributed to GSD.
 * Alexander Stukowski, OVITO GmbH
 * Charlotte Shiqi Zhao, University of Michigan
 * Tim Moore, University of Michigan
-* Joseph Burkhart, University of Michigan
\ No newline at end of file
+* Joseph Burkhart, University of Michigan

From 34229a1400aed19a01e7d83ecde5712f55c23770 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Fri, 20 Sep 2024 10:03:30 -0400
Subject: [PATCH 23/42] Fix clang tidy error

---
 gsd/gsd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index bd9de68b..bd13bc2e 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1618,7 +1618,7 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
     if ((handle->open_flags == GSD_OPEN_READWRITE || handle->open_flags == GSD_OPEN_APPEND)
         && (handle->header.gsd_version
             != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
-        && (handle->header.gsd_version >> 16 == GSD_CURRENT_FILE_VERSION_MAJOR))
+        && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR))
         {
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);

From 0a64636c4b9722c222b6a3f52d1157e281761fd0 Mon Sep 17 00:00:00 2001
From: Joseph Burkhart <61951318+josephburkhart@users.noreply.github.com>
Date: Tue, 24 Sep 2024 12:44:59 -0400
Subject: [PATCH 24/42] Remove commented code

Co-authored-by: Joshua A. Anderson <joaander@umich.edu>
---
 gsd/fl.pyx | 15 ---------------
 1 file changed, 15 deletions(-)

diff --git a/gsd/fl.pyx b/gsd/fl.pyx
index cb9bbc16..96c19435 100644
--- a/gsd/fl.pyx
+++ b/gsd/fl.pyx
@@ -146,21 +146,6 @@ cdef void * __get_ptr_float64(data):
     else:
         return <void*>&data_array_float64[0, 0]
 
-cdef void * __get_ptr_char(data):
-    # cdef numpy.ndarray[numpy.str_, ndim=2, mode="c", cast=True] data_array_char
-    cdef size_t address
-    address = data.__array_interface__["data"][0]
-    # data_array_char: cython.char[::1] = data.reshape(data.size)
-    # data_array_char = data
-    # cdef char *data_array_char
-    # data_array_char = cython.address(data.data[0])
-    if (data.size == 0):
-        return NULL
-    else:
-        # return <void*>cython.address(data_array_char.data[0])
-        # return <void*>cython.address(data_array_char[0])
-        return <void*>address
-
 
 def open(name, mode, application=None, schema=None, schema_version=None):
     """open(name, mode, application=None, schema=None, schema_version=None)

From 10e96ea734bade2397787c3512dd6e1ef84dd838 Mon Sep 17 00:00:00 2001
From: Joseph Burkhart <61951318+josephburkhart@users.noreply.github.com>
Date: Tue, 24 Sep 2024 12:45:13 -0400
Subject: [PATCH 25/42] Remove commented code

Co-authored-by: Joshua A. Anderson <joaander@umich.edu>
---
 gsd/fl.pyx | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/gsd/fl.pyx b/gsd/fl.pyx
index 96c19435..853cb0d0 100644
--- a/gsd/fl.pyx
+++ b/gsd/fl.pyx
@@ -625,12 +625,6 @@ cdef class GSDFile:
             elif data_array.dtype == numpy.float64:
                 gsd_type = libgsd.GSD_TYPE_DOUBLE
                 data_ptr = __get_ptr_float64(data_array)
-            # elif data_array.dtype.type is numpy.str_:
-            #     if N != 0:
-            #         M = int(data_array.dtype.itemsize/4)
-            #     print(f"Wrote: {N=}, {M=}, {data_array=}")
-            #     gsd_type = libgsd.GSD_TYPE_CHARACTER
-            #     data_ptr = __get_ptr_char(data_array)
             else:
                 raise ValueError("invalid type for chunk: " + name)
 

From 899e0e21f772122b433234bcc54898662b112204 Mon Sep 17 00:00:00 2001
From: Joseph Burkhart <61951318+josephburkhart@users.noreply.github.com>
Date: Tue, 24 Sep 2024 12:45:35 -0400
Subject: [PATCH 26/42] Remove unused import

Co-authored-by: Joshua A. Anderson <joaander@umich.edu>
---
 gsd/fl.pyx | 1 -
 1 file changed, 1 deletion(-)

diff --git a/gsd/fl.pyx b/gsd/fl.pyx
index 853cb0d0..e8f42ec5 100644
--- a/gsd/fl.pyx
+++ b/gsd/fl.pyx
@@ -22,7 +22,6 @@ from libc.stdint cimport uint8_t, int8_t, uint16_t, int16_t, uint32_t, int32_t,\
 from libc.errno cimport errno
 cimport gsd.libgsd as libgsd
 cimport numpy
-cimport cython
 
 logger = logging.getLogger('gsd.fl')
 

From 983f639aade0169e1834ed7f02771fdb0d054647 Mon Sep 17 00:00:00 2001
From: Joseph Burkhart <61951318+josephburkhart@users.noreply.github.com>
Date: Tue, 24 Sep 2024 12:46:07 -0400
Subject: [PATCH 27/42] Remove print statement

Co-authored-by: Joshua A. Anderson <joaander@umich.edu>
---
 gsd/fl.pyx | 1 -
 1 file changed, 1 deletion(-)

diff --git a/gsd/fl.pyx b/gsd/fl.pyx
index e8f42ec5..bc65046d 100644
--- a/gsd/fl.pyx
+++ b/gsd/fl.pyx
@@ -853,7 +853,6 @@ cdef class GSDFile:
             if gsd_type == libgsd.GSD_TYPE_CHARACTER:
                 data_array = data_array.flatten()
                 bytes_array = data_array.view(dtype=numpy.dtype((bytes, data_array.shape[0])))
-                print(bytes_array[0].decode("UTF-8"))
                 return bytes_array[0].decode("UTF-8")
 
             return data_array.reshape([index_entry.N])

From 900f43faab210b1275a8cbc244b4a3e79a1bbf93 Mon Sep 17 00:00:00 2001
From: Joseph Burkhart <61951318+josephburkhart@users.noreply.github.com>
Date: Tue, 24 Sep 2024 12:47:12 -0400
Subject: [PATCH 28/42] Fix header overwriting when file version reflects
 current version

Co-authored-by: Joshua A. Anderson <joaander@umich.edu>
---
 gsd/gsd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index bd13bc2e..5d2c77b7 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1618,7 +1618,8 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
     if ((handle->open_flags == GSD_OPEN_READWRITE || handle->open_flags == GSD_OPEN_APPEND)
         && (handle->header.gsd_version
             != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
-        && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR))
+        && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR)
+        && (handle->header.gsd_version & 0xFFFF < GSD_CURRENT_FILE_VERSION_MINOR))
         {
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);

From 74731b00a4234b0dd22ae00f1a56cb037a27d96c Mon Sep 17 00:00:00 2001
From: Joseph Burkhart <61951318+josephburkhart@users.noreply.github.com>
Date: Tue, 24 Sep 2024 12:47:41 -0400
Subject: [PATCH 29/42] Revise comment for clarity

Co-authored-by: Joshua A. Anderson <joaander@umich.edu>
---
 gsd/gsd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index 5d2c77b7..c26e4138 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -2583,7 +2583,7 @@ int gsd_upgrade(struct gsd_handle* handle)
                 }
             }
 
-        // label the file as a v2.1 file
+        // GSD always writes files matching the current major and minor version.
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);
 

From f84a50767de840a0710fce1e419d9fdffbf7744b Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Tue, 24 Sep 2024 12:58:07 -0400
Subject: [PATCH 30/42] Fix clang tidy error

---
 gsd/gsd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index c26e4138..d0e412e5 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1619,7 +1619,7 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
         && (handle->header.gsd_version
             != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
         && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR)
-        && (handle->header.gsd_version & 0xFFFF < GSD_CURRENT_FILE_VERSION_MINOR))
+        && (handle->header.gsd_version & (0xFFFF < GSD_CURRENT_FILE_VERSION_MINOR)))
         {
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);

From 377c2debcef43a463f2e96f548b708598786f081 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Tue, 24 Sep 2024 13:09:29 -0400
Subject: [PATCH 31/42] Fix clang tidy magic number error

---
 gsd/gsd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index d0e412e5..2f6731b7 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1615,11 +1615,12 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
 
     // When opening a file in a writeable mode, if the file's major version
     // is identical to the current major version, silently upgrade the minor version
+    const unsigned int GSD_MAX_FILE_VERSION_MINOR = 0xFFFF;
     if ((handle->open_flags == GSD_OPEN_READWRITE || handle->open_flags == GSD_OPEN_APPEND)
         && (handle->header.gsd_version
             != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
         && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR)
-        && (handle->header.gsd_version & (0xFFFF < GSD_CURRENT_FILE_VERSION_MINOR)))
+        && (handle->header.gsd_version & (GSD_MAX_FILE_VERSION_MINOR < GSD_CURRENT_FILE_VERSION_MINOR)))
         {
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);

From a01f29b5beb1eadde9e4a41ed55659ce0bbc41a2 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Tue, 24 Sep 2024 13:10:46 -0400
Subject: [PATCH 32/42] Fix formatting error with pre-commit

---
 gsd/gsd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index 2f6731b7..af220719 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1620,7 +1620,8 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
         && (handle->header.gsd_version
             != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
         && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR)
-        && (handle->header.gsd_version & (GSD_MAX_FILE_VERSION_MINOR < GSD_CURRENT_FILE_VERSION_MINOR)))
+        && (handle->header.gsd_version
+            & (GSD_MAX_FILE_VERSION_MINOR < GSD_CURRENT_FILE_VERSION_MINOR)))
         {
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);

From ffe16521a86540153c97f4d0a14746c9f7a2bef0 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Wed, 25 Sep 2024 10:12:40 -0400
Subject: [PATCH 33/42] Remove unneeded check.

---
 gsd/gsd.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/gsd/gsd.c b/gsd/gsd.c
index af220719..07997408 100644
--- a/gsd/gsd.c
+++ b/gsd/gsd.c
@@ -1613,15 +1613,12 @@ inline static int gsd_initialize_handle(struct gsd_handle* handle)
     handle->maximum_write_buffer_size = GSD_DEFAULT_MAXIMUM_WRITE_BUFFER_SIZE;
     handle->index_entries_to_buffer = GSD_DEFAULT_INDEX_ENTRIES_TO_BUFFER;
 
-    // When opening a file in a writeable mode, if the file's major version
-    // is identical to the current major version, silently upgrade the minor version
-    const unsigned int GSD_MAX_FILE_VERSION_MINOR = 0xFFFF;
+    // Silently upgrade writable files from a previous matching major version to the latest
+    // minor version.
     if ((handle->open_flags == GSD_OPEN_READWRITE || handle->open_flags == GSD_OPEN_APPEND)
         && (handle->header.gsd_version
             != gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR))
-        && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR)
-        && (handle->header.gsd_version
-            & (GSD_MAX_FILE_VERSION_MINOR < GSD_CURRENT_FILE_VERSION_MINOR)))
+        && (handle->header.gsd_version >> (sizeof(uint32_t) * 4) == GSD_CURRENT_FILE_VERSION_MAJOR))
         {
         handle->header.gsd_version
             = gsd_make_version(GSD_CURRENT_FILE_VERSION_MAJOR, GSD_CURRENT_FILE_VERSION_MINOR);

From 6f37b8ffefaa8cead6b7ee5433a4ef8348f0fb79 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Wed, 25 Sep 2024 10:26:32 -0400
Subject: [PATCH 34/42] Reword GSD_CHARACTER definition in docs.

---
 doc/file-layer.rst | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/doc/file-layer.rst b/doc/file-layer.rst
index 96c1b50d..8bec0f8a 100644
--- a/doc/file-layer.rst
+++ b/doc/file-layer.rst
@@ -6,7 +6,7 @@ File layer
 
 .. highlight:: c
 
-**Version: 2.0**
+**Version: 2.x**
 
 General simulation data (GSD) **file layer** design and rationale. These use
 cases and design specifications define the low level GSD file format.
@@ -236,5 +236,8 @@ A data block stores raw data bytes on the disk. For a given index entry
 ``entry``, the data starts at location ``entry.location`` and is the next
 ``entry.N * entry.M * gsd_sizeof_type(entry.type)`` bytes.
 
-The type fields are self-explanatory except for UTF-8 strings, for which
-null-termination is allowed but not required.
+Added in version 2.1
+^^^^^^^^^^^^^^^^^^^^
+
+* The ``GSD_CHARACTER`` chunk type represents a UTF-8 string (null termination is allowed, but not
+  required).

From 1ec21b9c9c5fb70c5b7d2bb452f5aa15e80c4cc3 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Wed, 25 Sep 2024 10:55:23 -0400
Subject: [PATCH 35/42] Further revisions to the file layer spec.

---
 doc/file-layer.rst | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/doc/file-layer.rst b/doc/file-layer.rst
index 8bec0f8a..a79fb1dc 100644
--- a/doc/file-layer.rst
+++ b/doc/file-layer.rst
@@ -128,7 +128,7 @@ There are four types of data blocks in a GSD file.
 
    * List of string names used by index entries.
    * v1.0 files: Each name is a 64-byte character string.
-   * v2.0 files: Names may have any length and are separated by 0 terminators.
+   * v2.x files: Names may have any length and are separated by 0 terminators.
    * The first name that starts with the 0 byte marks the end of the list
    * The header stores the total size of the name list block.
 
@@ -215,13 +215,13 @@ non-standard packing attributes or pragmas to enforce this.
 In v1.0 files, the frame index must monotonically increase from one index entry
 to the next. The GSD API ensures this.
 
-In v2.0 files, the entire index block is stored sorted first by frame, then
+In v2.x files, the entire index block is stored sorted first by frame, then
 by *id*.
 
 Namelist block
 ^^^^^^^^^^^^^^
 
-In v2.0 files, the namelist block stores a list of strings separated by 0
+In v2.x files, the namelist block stores a list of strings separated by 0
 terminators.
 
 In v1.0 files, the namelist block stores a list of 0-terminated strings in
@@ -237,7 +237,7 @@ A data block stores raw data bytes on the disk. For a given index entry
 ``entry.N * entry.M * gsd_sizeof_type(entry.type)`` bytes.
 
 Added in version 2.1
-^^^^^^^^^^^^^^^^^^^^
+--------------------
 
 * The ``GSD_CHARACTER`` chunk type represents a UTF-8 string (null termination is allowed, but not
   required).

From 3e2ae056d1ba939bbca42c1c996c1b703d02f869 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Wed, 25 Sep 2024 11:41:29 -0400
Subject: [PATCH 36/42] Update string writing example.

---
 doc/fl-examples.rst | 14 ++++----------
 1 file changed, 4 insertions(+), 10 deletions(-)

diff --git a/doc/fl-examples.rst b/doc/fl-examples.rst
index d66a47e3..e61b84c3 100644
--- a/doc/fl-examples.rst
+++ b/doc/fl-examples.rst
@@ -198,21 +198,15 @@ Store string chunks
                     application="My application",
                     schema="My Schema",
                     schema_version=[1,0])
-    f.mode
-    s = "This is a string"
-    b = numpy.array([s], dtype=numpy.dtype((bytes, len(s)+1)))
-    b = b.view(dtype=numpy.int8)
-    b
-    f.write_chunk(name='string', data=b)
+    f.write_chunk(name='string', data="This is a string")
     f.end_frame()
     r = f.read_chunk(frame=0, name='string')
     r
-    r = r.view(dtype=numpy.dtype((bytes, r.shape[0])));
-    r[0].decode('UTF-8')
     f.close()
 
-To store a string in a gsd file, convert it to a numpy array of bytes and store
-that data in the file. Decode the byte sequence to get back a string.
+Staring with GSD 3.4.0, the file layer can natively store strings in the file.
+In previous versions, you need to convert strings to a numpy array of bytes and store
+that data in the file.
 
 Truncate
 ^^^^^^^^

From e1bb00317178df2f29263de7ae8bd41ca2670834 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Thu, 26 Sep 2024 10:10:34 -0400
Subject: [PATCH 37/42] Tweak log reading to correctly duck-type strings

---
 gsd/hoomd.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/gsd/hoomd.py b/gsd/hoomd.py
index 7af28a29..a4093339 100644
--- a/gsd/hoomd.py
+++ b/gsd/hoomd.py
@@ -1195,6 +1195,9 @@ def read_log(name, scalar_only=False):
                     tmp = numpy.array([0], dtype=numpy.uint64)
                 else:
                     tmp = gsdfileobj.read_chunk(frame=0, name=log)
+                    # if chunk contains string, put it in the numpy array
+                    if isinstance(tmp, str):
+                        tmp = numpy.array([tmp])
 
                 if scalar_only and not tmp.shape[0] == 1:
                     continue

From 3c5ef91cbb35f584e1ddafdccb1c9e1b937febe1 Mon Sep 17 00:00:00 2001
From: josephburkhart <joaburkh@umich.edu>
Date: Thu, 26 Sep 2024 10:10:51 -0400
Subject: [PATCH 38/42] Add string data type to log tests

---
 gsd/test/test_hoomd.py | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

diff --git a/gsd/test/test_hoomd.py b/gsd/test/test_hoomd.py
index 169180bb..eb4ca65c 100644
--- a/gsd/test/test_hoomd.py
+++ b/gsd/test/test_hoomd.py
@@ -794,11 +794,13 @@ def test_log(tmp_path, open_mode):
     frame0.log['particles/pair_lj_energy'] = [0, -5, -8, -3]
     frame0.log['value/potential_energy'] = [10]
     frame0.log['value/pressure'] = [-3]
+    frame0.log['category'] = 'A'
 
     frame1 = gsd.hoomd.Frame()
 
     frame1.log['particles/pair_lj_energy'] = [1, 2, -4, -10]
     frame1.log['value/pressure'] = [5]
+    frame1.log['category'] = 'B'
 
     with gsd.hoomd.open(name=tmp_path / 'test_log.gsd', mode=open_mode.write) as hf:
         hf.extend([frame0, frame1])
@@ -819,6 +821,7 @@ def test_log(tmp_path, open_mode):
         numpy.testing.assert_array_equal(
             s.log['value/pressure'], frame0.log['value/pressure']
         )
+        assert s.log['category'] == frame0.log['category']
 
         s = hf[1]
 
@@ -837,6 +840,7 @@ def test_log(tmp_path, open_mode):
         numpy.testing.assert_array_equal(
             s.log['value/pressure'], frame1.log['value/pressure']
         )
+        assert s.log['category'] == frame1.log['category']
 
 
 def test_pickle(tmp_path):
@@ -877,6 +881,7 @@ def test_read_log(tmp_path):
     ]
     frame0.log['value/potential_energy'] = [10]
     frame0.log['value/pressure'] = [-3]
+    frame0.log['category'] = 'A'
 
     frame1 = gsd.hoomd.Frame()
     frame1.configuration.step = 1
@@ -888,6 +893,7 @@ def test_read_log(tmp_path):
         (4, 4, 4),
     ]
     frame1.log['value/pressure'] = [5]
+    frame1.log['category'] = 'B'
 
     with gsd.hoomd.open(name=tmp_path / 'test_log.gsd', mode='w') as hf:
         hf.extend([frame0, frame1])
@@ -897,13 +903,14 @@ def test_read_log(tmp_path):
         name=tmp_path / 'test_log.gsd', scalar_only=False
     )
 
-    assert len(logged_data_dict) == 5
+    assert len(logged_data_dict) == 6
     assert list(logged_data_dict.keys()) == [
         'configuration/step',
         'log/particles/pair_lj_energy',
         'log/particles/pair_lj_force',
         'log/value/potential_energy',
         'log/value/pressure',
+        'log/category',
     ]
 
     numpy.testing.assert_array_equal(logged_data_dict['configuration/step'], [0, 1])
@@ -926,16 +933,21 @@ def test_read_log(tmp_path):
         logged_data_dict['log/value/pressure'],
         [*frame0.log['value/pressure'], *frame1.log['value/pressure']],
     )
+    numpy.testing.assert_array_equal(
+        logged_data_dict['log/category'],
+        [*frame0.log['category'], *frame1.log['category']],
+    )
 
     # Test scalar_only = True
     logged_data_dict = gsd.hoomd.read_log(
         name=tmp_path / 'test_log.gsd', scalar_only=True
     )
-    assert len(logged_data_dict) == 3
+    assert len(logged_data_dict) == 4
     assert list(logged_data_dict.keys()) == [
         'configuration/step',
         'log/value/potential_energy',
         'log/value/pressure',
+        'log/category',
     ]
     numpy.testing.assert_array_equal(logged_data_dict['configuration/step'], [0, 1])
     numpy.testing.assert_array_equal(
@@ -946,6 +958,10 @@ def test_read_log(tmp_path):
         logged_data_dict['log/value/pressure'],
         [*frame0.log['value/pressure'], *frame1.log['value/pressure']],
     )
+    numpy.testing.assert_array_equal(
+        logged_data_dict['log/category'],
+        [*frame0.log['category'], *frame1.log['category']],
+    )
 
 
 def test_read_log_warning(tmp_path):

From eca3c6a40889bbf66151fd65d495f9a5748c3fef Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Fri, 18 Oct 2024 09:07:59 -0400
Subject: [PATCH 39/42] Use StringDType in read_log for string chunks.

---
 gsd/hoomd.py           | 13 ++++++++++---
 gsd/test/test_hoomd.py | 14 ++++++++++----
 2 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/gsd/hoomd.py b/gsd/hoomd.py
index a4093339..3934a8f3 100644
--- a/gsd/hoomd.py
+++ b/gsd/hoomd.py
@@ -1197,13 +1197,15 @@ def read_log(name, scalar_only=False):
                     tmp = gsdfileobj.read_chunk(frame=0, name=log)
                     # if chunk contains string, put it in the numpy array
                     if isinstance(tmp, str):
-                        tmp = numpy.array([tmp])
+                        tmp = numpy.array([tmp], dtype=numpy.dtypes.StringDType)
 
                 if scalar_only and not tmp.shape[0] == 1:
                     continue
                 if tmp.shape[0] == 1:
                     logged_data_dict[log] = numpy.full(
-                        fill_value=tmp[0], shape=(gsdfileobj.nframes,)
+                        fill_value=tmp[0],
+                        shape=(gsdfileobj.nframes,),
+                        dtype=tmp.dtype,
                     )
                 else:
                     logged_data_dict[log] = numpy.tile(
@@ -1215,7 +1217,12 @@ def read_log(name, scalar_only=False):
                     if not gsdfileobj.chunk_exists(frame=idx, name=key):
                         continue
                     data = gsdfileobj.read_chunk(frame=idx, name=key)
-                    if len(logged_data_dict[key][idx].shape) == 0:
+                    if (
+                        not isinstance(
+                            logged_data_dict[key].dtype, numpy.dtypes.StringDType
+                        )
+                        and len(logged_data_dict[key][idx].shape) == 0
+                    ):
                         logged_data_dict[key][idx] = data[0]
                     else:
                         logged_data_dict[key][idx] = data
diff --git a/gsd/test/test_hoomd.py b/gsd/test/test_hoomd.py
index eb4ca65c..3b813c17 100644
--- a/gsd/test/test_hoomd.py
+++ b/gsd/test/test_hoomd.py
@@ -800,7 +800,7 @@ def test_log(tmp_path, open_mode):
 
     frame1.log['particles/pair_lj_energy'] = [1, 2, -4, -10]
     frame1.log['value/pressure'] = [5]
-    frame1.log['category'] = 'B'
+    frame1.log['category'] = 'BBB'
 
     with gsd.hoomd.open(name=tmp_path / 'test_log.gsd', mode=open_mode.write) as hf:
         hf.extend([frame0, frame1])
@@ -893,7 +893,7 @@ def test_read_log(tmp_path):
         (4, 4, 4),
     ]
     frame1.log['value/pressure'] = [5]
-    frame1.log['category'] = 'B'
+    frame1.log['category'] = 'BBB'
 
     with gsd.hoomd.open(name=tmp_path / 'test_log.gsd', mode='w') as hf:
         hf.extend([frame0, frame1])
@@ -935,7 +935,10 @@ def test_read_log(tmp_path):
     )
     numpy.testing.assert_array_equal(
         logged_data_dict['log/category'],
-        [*frame0.log['category'], *frame1.log['category']],
+        numpy.array(
+            [frame0.log['category'], frame1.log['category']],
+            dtype=numpy.dtypes.StringDType,
+        ),
     )
 
     # Test scalar_only = True
@@ -960,7 +963,10 @@ def test_read_log(tmp_path):
     )
     numpy.testing.assert_array_equal(
         logged_data_dict['log/category'],
-        [*frame0.log['category'], *frame1.log['category']],
+        numpy.array(
+            [frame0.log['category'], frame1.log['category']],
+            dtype=numpy.dtypes.StringDType,
+        ),
     )
 
 

From 50c8b0b7dcd0469bf92ae4183d53a79b97612f76 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Fri, 18 Oct 2024 09:25:28 -0400
Subject: [PATCH 40/42] Require numpy 2.0.

---
 .github/workflows/build_wheels.yaml | 8 ++++----
 INSTALLING.rst                      | 2 +-
 pyproject.toml                      | 2 +-
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/build_wheels.yaml b/.github/workflows/build_wheels.yaml
index 4fbe2ddb..b847e5ca 100644
--- a/.github/workflows/build_wheels.yaml
+++ b/.github/workflows/build_wheels.yaml
@@ -38,11 +38,11 @@ jobs:
 
         python:
         - version: 'cp310'
-          oldest_numpy: '1.21.6'
+          oldest_numpy: '2.0.0'
         - version: 'cp311'
-          oldest_numpy: '1.23.2'
+          oldest_numpy: '2.0.0'
         - version: 'cp312'
-          oldest_numpy: '1.26.2'
+          oldest_numpy: '2.0.0'
         - version: 'cp313'
           oldest_numpy: '2.1.1'
 
@@ -53,7 +53,7 @@ jobs:
         uses: pypa/cibuildwheel@d4a2945fcc8d13f20a1b99d461b8e844d5fc6e23 # v2.21.1
         env:
           CIBW_BUILD: "${{ matrix.python.version }}-*"
-          CIBW_TEST_REQUIRES: pytest==8.2.1 numpy==${{ matrix.python.oldest_numpy }}
+          CIBW_TEST_REQUIRES: pytest==8.3.3 numpy==${{ matrix.python.oldest_numpy }}
 
       - uses: actions/upload-artifact@50769540e7f4bd5e21e526ee35c689e35e0d6874 # v4.4.0
         with:
diff --git a/INSTALLING.rst b/INSTALLING.rst
index 712ac972..af2f9472 100644
--- a/INSTALLING.rst
+++ b/INSTALLING.rst
@@ -109,7 +109,7 @@ Install prerequisites
 
 * **C compiler** (tested with gcc 10-14, clang 10-18, Visual Studio 2019-2022)
 * **Python** >= 3.10
-* **numpy** >= 1.19.0
+* **numpy** >= 2.0.0
 * **Cython** >= 0.22
 
 **To execute unit tests:**
diff --git a/pyproject.toml b/pyproject.toml
index dfb43604..700919f0 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -17,7 +17,7 @@ classifiers=[
     "License :: OSI Approved :: BSD License",
     "Topic :: Scientific/Engineering :: Physics",
     ]
-dependencies = ["numpy>=1.19.0"]
+dependencies = ["numpy>=2.0.0"]
 
 [project.scripts]
 gsd = "gsd.__main__:main"

From 12e0ea63d23d21a465d2cf9ce56f141a30a58e07 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Fri, 18 Oct 2024 09:31:34 -0400
Subject: [PATCH 41/42] Test with Python 3.13 final.

---
 .github/workflows/unit_test.yaml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/unit_test.yaml b/.github/workflows/unit_test.yaml
index f0d0d43a..e9b8da77 100644
--- a/.github/workflows/unit_test.yaml
+++ b/.github/workflows/unit_test.yaml
@@ -30,7 +30,7 @@ jobs:
         - os: windows-2019
           python: '3.12'
         - os: windows-2022
-          python: '3.13.0-rc.1'
+          python: '3.13'
         ##############
         # Mac
         # macos-x86_64
@@ -38,7 +38,7 @@ jobs:
           python: '3.12'
         # macos-arm64
         - os: macos-14
-          python: '3.13.0-rc.1'
+          python: '3.13'
         ##############
         # Ubuntu 24.04
         - os: ubuntu-24.04
@@ -58,7 +58,7 @@ jobs:
           c_compiler: clang-18
           cxx_compiler: clang++-18
         - os: ubuntu-24.04
-          python: '3.13.0-rc.1'
+          python: '3.13.0'
           c_compiler: clang-18
           cxx_compiler: clang++-18
         ##############

From 59e421385ba4d9156a828252720d6a8f31e0dfa6 Mon Sep 17 00:00:00 2001
From: "Joshua A. Anderson" <joaander@umich.edu>
Date: Fri, 18 Oct 2024 09:32:59 -0400
Subject: [PATCH 42/42] Mention numpy 2.0 requirement in change log.

---
 CHANGELOG.rst | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index 70523a6e..e0a0636e 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -18,6 +18,11 @@ Change Log
 * New chunk type for string data - valid in file layer versions 2.1 and later
   (`#391 <https://github.com/glotzerlab/gsd/pull/391>`__).
 
+*Changed:*
+
+* Require NumPy >= 2.0
+  (`#391 <https://github.com/glotzerlab/gsd/pull/391>`__).
+
 3.3.2 (2024-09-06)
 ^^^^^^^^^^^^^^^^^^