From 0953befb34d6f1683b47b064877df46e2ede7678 Mon Sep 17 00:00:00 2001
From: Peter Williams <peter@newton.cx>
Date: Tue, 13 Sep 2022 17:33:23 -0400
Subject: [PATCH 1/4] wwt_data_formats/plate.py: add this module

Skeleton support for dealing with plate files. It turns out that the 4
GB plate file format (which I'll call "V1") is incredibly easy to
handle, so that's nice.
---
 .../wwt_data_formats.plate.V1PlateReader.rst  | 19 ++++
 docs/api/wwt_data_formats.plate.rst           |  3 +
 docs/index.rst                                |  1 +
 wwt_data_formats/plate.py                     | 95 +++++++++++++++++++
 4 files changed, 118 insertions(+)
 create mode 100644 docs/api/wwt_data_formats.plate.V1PlateReader.rst
 create mode 100644 docs/api/wwt_data_formats.plate.rst
 create mode 100644 wwt_data_formats/plate.py

diff --git a/docs/api/wwt_data_formats.plate.V1PlateReader.rst b/docs/api/wwt_data_formats.plate.V1PlateReader.rst
new file mode 100644
index 0000000..c4dab44
--- /dev/null
+++ b/docs/api/wwt_data_formats.plate.V1PlateReader.rst
@@ -0,0 +1,19 @@
+V1PlateReader
+=============
+
+.. currentmodule:: wwt_data_formats.plate
+
+.. autoclass:: V1PlateReader
+   :show-inheritance:
+
+   .. rubric:: Methods Summary
+
+   .. autosummary::
+
+      ~V1PlateReader.close
+      ~V1PlateReader.read_tile
+
+   .. rubric:: Methods Documentation
+
+   .. automethod:: close
+   .. automethod:: read_tile
diff --git a/docs/api/wwt_data_formats.plate.rst b/docs/api/wwt_data_formats.plate.rst
new file mode 100644
index 0000000..be3a0d4
--- /dev/null
+++ b/docs/api/wwt_data_formats.plate.rst
@@ -0,0 +1,3 @@
+.. automodapi:: wwt_data_formats.plate
+   :no-inheritance-diagram:
+   :inherited-members:
diff --git a/docs/index.rst b/docs/index.rst
index 00ddef0..83a8579 100644
--- a/docs/index.rst
+++ b/docs/index.rst
@@ -46,6 +46,7 @@ API Reference
    api/wwt_data_formats.imageset
    api/wwt_data_formats.layers
    api/wwt_data_formats.place
+   api/wwt_data_formats.plate
    api/wwt_data_formats.server
 
 
diff --git a/wwt_data_formats/plate.py b/wwt_data_formats/plate.py
new file mode 100644
index 0000000..f09cc32
--- /dev/null
+++ b/wwt_data_formats/plate.py
@@ -0,0 +1,95 @@
+# -*- mode: python; coding: utf-8 -*-
+# Copyright 2022 the .NET Foundation
+# Licensed under the MIT License.
+
+"""The "plate" container formats for binary files.
+
+"Plate files" are used in some parts of WWT to assemble large numbers of small
+files into one big file. In particular, most of the original core WWT tile data
+assets are compiled into plate files.
+
+There are three variants of the plate file format. The oldest variant (let's
+call it V0) needs external information in order to be read correctly and is not
+yet supported here.
+
+The next version (let's call it V1) has an eight-byte header followed by all
+file location information in a fixed structure. Offsets are 32 bits, and so the
+total file size is limited to 4 GiB. The densely-populated header is not
+efficient for sparsely-populated tile pyramids. The reference implementation for
+the V1 format is in `PlateTilePyramid.cs`_.
+
+.. _PlateTilePyramid.cs: https://github.com/WorldWideTelescope/wwt-website/blob/master/src/WWT.PlateFiles/PlateTilePyramid.cs
+
+The last version (V2) uses a hash table format. The V2 format is more efficient
+for sparsely-populated tile pyramids, and supports files that are (much) larger
+than 4 GiB. The reference implementation for the V2 format is in
+`PlateFile2.cs`_. The V2 format is used by WWT's HiRISE data.
+
+.. _PlateFile2.cs: https://github.com/WorldWideTelescope/wwt-website/blob/master/src/WWT.PlateFiles/PlateFile2.cs
+
+"""
+
+__all__ = """
+V1PlateReader
+""".split()
+
+from struct import unpack
+
+
+class V1PlateReader(object):
+    """Reader for the "V1" plate file format.
+
+    Unlike most of the other WWT data formats implemented in this package,
+    plate files are stored in a simple binary structure, not XML."""
+
+    _stream = None
+    _levels: int
+
+    def __init__(self, stream):
+        self._stream = stream
+
+        # We must have random access to the stream.
+        stream.seek(0)
+        magic, levels = unpack("<II", stream.read(8))
+
+        if magic != 0x7E69AD43:
+            if magic == 0x17914242:
+                raise Exception("input stream is a V2 plate file, not V1")
+            raise Exception("input stream does not look like a V1 plate file (nor V2)")
+
+        self._levels = levels
+
+    def close(self):
+        """Close the underlying stream, making this object essentially unusable."""
+        self._stream.close()
+        self._stream = None
+
+    def read_tile(self, level, x, y):
+        """Read the specified tile position into memory in its entirety and
+        return its contents.
+
+        Returns bytes."""
+
+        if self._stream is None:
+            raise Exception("cannot read a closed FileCabinetReader")
+
+        if level < 0 or level > self._levels:
+            raise ValueError(f"invalid `level` {level}")
+
+        n = 2**level
+
+        if x < 0 or x >= n or y < 0 or y >= n:
+            raise ValueError(f"invalid tile position L{level}X{x}Y{y}")
+
+        # This is the total number of tiles in all levels from 0 to `level - 1`,
+        # plus one to account for the header item:
+        index = (4**level - 1) // 3 + 1
+
+        # The offset of this tile within the level:
+        index += n * y + x
+
+        self._stream.seek(8 * index)
+        offset, length = unpack("<II", self._stream.read(8))
+
+        self._stream.seek(offset)
+        return self._stream.read(length)

From f75ec50e8179c0209ecba0f3a69b0ca62bc562e4 Mon Sep 17 00:00:00 2001
From: Peter Williams <peter@newton.cx>
Date: Wed, 14 Sep 2022 15:15:20 -0400
Subject: [PATCH 2/4] wwt_data_formats/tests/test_folder.py: use centralized
 text fixture

---
 wwt_data_formats/tests/test_folder.py | 26 ++++----------------------
 1 file changed, 4 insertions(+), 22 deletions(-)

diff --git a/wwt_data_formats/tests/test_folder.py b/wwt_data_formats/tests/test_folder.py
index 3eb3b91..315a418 100644
--- a/wwt_data_formats/tests/test_folder.py
+++ b/wwt_data_formats/tests/test_folder.py
@@ -6,31 +6,13 @@ from __future__ import absolute_import, division, print_function
 
 from mock import Mock
 import os.path
-import shutil
-import tempfile
 import pytest
 from xml.etree import ElementTree as etree
 
-from . import assert_xml_trees_equal, test_path
+from . import assert_xml_trees_equal, tempdir, test_path, work_in_tempdir
 from .. import cli, folder, imageset, place
 
 
-@pytest.fixture
-def tempdir():
-    d = tempfile.mkdtemp()
-    yield d
-    shutil.rmtree(d)
-
-
-@pytest.fixture
-def in_tempdir(tempdir):
-    prev_dir = os.getcwd()
-    os.chdir(tempdir)
-    yield tempdir
-    # Windows can't remove the temp tree unless we chdir out of it.
-    os.chdir(prev_dir)
-
-
 BASIC_XML_STRING = """
 <Folder Browseable="True" Group="Explorer" Searchable="True" />
 """
@@ -205,7 +187,7 @@ def test_wtml_report():
     cli.entrypoint(["wtml", "report", test_path("report_rel.wtml")])
 
 
-def test_wtml_rewrite_disk(in_tempdir):
+def test_wtml_rewrite_disk(work_in_tempdir):
     f = folder.Folder()
     f.url = "sub%20dir/image.jpg"
 
@@ -215,12 +197,12 @@ def test_wtml_rewrite_disk(in_tempdir):
     cli.entrypoint(["wtml", "rewrite-disk", "index_rel.wtml", "index_disk.wtml"])
 
     f = folder.Folder.from_file("index_disk.wtml")
-    # abspath('') is not necessarily equal to abspath(in_tempdir), due to
+    # abspath('') is not necessarily equal to abspath(work_in_tempdir), due to
     # symlinks and Windows filename shorterning.
     assert f.url == os.path.join(os.path.abspath(""), "sub dir", "image.jpg")
 
 
-def test_wtml_rewrite_urls(in_tempdir):
+def test_wtml_rewrite_urls(work_in_tempdir):
     f = folder.Folder()
     f.url = "../updir/somewhere.wtml"
 

From 430a77f60aea9908dc013f2f27fefce14e90a9aa Mon Sep 17 00:00:00 2001
From: Peter Williams <peter@newton.cx>
Date: Wed, 14 Sep 2022 15:40:59 -0400
Subject: [PATCH 3/4] wwt_data_formats/plate.py: add writer, tests, docs

---
 .../wwt_data_formats.plate.V1PlateWriter.rst  |  21 ++
 wwt_data_formats/plate.py                     | 224 +++++++++++++++++-
 wwt_data_formats/tests/test_plate.py          |  50 ++++
 3 files changed, 284 insertions(+), 11 deletions(-)
 create mode 100644 docs/api/wwt_data_formats.plate.V1PlateWriter.rst
 create mode 100644 wwt_data_formats/tests/test_plate.py

diff --git a/docs/api/wwt_data_formats.plate.V1PlateWriter.rst b/docs/api/wwt_data_formats.plate.V1PlateWriter.rst
new file mode 100644
index 0000000..db7bcd3
--- /dev/null
+++ b/docs/api/wwt_data_formats.plate.V1PlateWriter.rst
@@ -0,0 +1,21 @@
+V1PlateWriter
+=============
+
+.. currentmodule:: wwt_data_formats.plate
+
+.. autoclass:: V1PlateWriter
+   :show-inheritance:
+
+   .. rubric:: Methods Summary
+
+   .. autosummary::
+
+      ~V1PlateWriter.append_bytes
+      ~V1PlateWriter.append_stream
+      ~V1PlateWriter.close
+
+   .. rubric:: Methods Documentation
+
+   .. automethod:: append_bytes
+   .. automethod:: append_stream
+   .. automethod:: close
diff --git a/wwt_data_formats/plate.py b/wwt_data_formats/plate.py
index f09cc32..63323ed 100644
--- a/wwt_data_formats/plate.py
+++ b/wwt_data_formats/plate.py
@@ -31,21 +31,32 @@ than 4 GiB. The reference implementation for the V2 format is in
 
 __all__ = """
 V1PlateReader
+V1PlateWriter
 """.split()
 
-from struct import unpack
+from io import BytesIO
+from struct import pack, unpack
+from typing import BinaryIO, List
 
 
 class V1PlateReader(object):
-    """Reader for the "V1" plate file format.
+    """Reader for the "V1" WWT plate file format.
 
-    Unlike most of the other WWT data formats implemented in this package,
-    plate files are stored in a simple binary structure, not XML."""
+    Parameters
+    ----------
+    stream : readable, seekable, bytes-based file-like object
+        The underlying data stream. If you explicitly :meth:`close` this object,
+        it will close the underlying stream.
 
-    _stream = None
+    Notes
+    -----
+    Unlike most of the other WWT data formats implemented in this package, plate
+    files are stored in a simple binary structure, not XML."""
+
+    _stream: BinaryIO
     _levels: int
 
-    def __init__(self, stream):
+    def __init__(self, stream: BinaryIO):
         self._stream = stream
 
         # We must have random access to the stream.
@@ -61,17 +72,37 @@ class V1PlateReader(object):
 
     def close(self):
         """Close the underlying stream, making this object essentially unusable."""
-        self._stream.close()
-        self._stream = None
+        if self._stream is not None:
+            self._stream.close()
+            self._stream = None
 
-    def read_tile(self, level, x, y):
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *_exc):
+        self.close()
+        return False
+
+    def read_tile(self, level: int, x: int, y: int) -> bytes:
         """Read the specified tile position into memory in its entirety and
         return its contents.
 
-        Returns bytes."""
+        Parameters
+        ----------
+        level : int
+            The level of the tile to read
+        x : int
+            The X position of the tile to read
+        y : int
+            The Y position of the tile to read
+
+        Returns
+        -------
+        data : bytes
+            The data for the specified tile position."""
 
         if self._stream is None:
-            raise Exception("cannot read a closed FileCabinetReader")
+            raise Exception("cannot read a closed V1PlateReader")
 
         if level < 0 or level > self._levels:
             raise ValueError(f"invalid `level` {level}")
@@ -93,3 +124,174 @@ class V1PlateReader(object):
 
         self._stream.seek(offset)
         return self._stream.read(length)
+
+
+class V1PlateWriter(object):
+    """Writer for the "V1" WWT plate file format.
+
+    Parameters
+    ----------
+    stream : writeable, seekable, bytes-based file-like object
+        The underlying data destination. This object becomes responsible for
+        closing the stream.
+    levels : int
+        The number of tile levels to allocate for this plate file. Must be
+        nonnegative.
+
+    Notes
+    -----
+    This file format assumes that tile data will be densely populated up to the
+    specified number of levels (although missing entries are allowed). The
+    maximum final file size is 4 GiB.
+
+    This object is usable as a context manager, and should be explicitly closed.
+
+    Unlike most of the other WWT data formats implemented in this package, plate
+    files are stored in a simple binary structure, not XML."""
+
+    _stream: BinaryIO
+    _levels: int
+    _next_offset: int
+    _filedata: List[bytes]
+
+    def __init__(self, stream: BinaryIO, levels: int):
+        self._stream = stream
+        self._levels = levels
+
+        if levels < 0:
+            raise ValueError(f"illegal `levels` value {levels!r}")
+
+        # We must have random access to the stream.
+        stream.seek(0)
+        stream.write(pack("<II", 0x7E69AD43, levels))
+
+        # Total number of tiles in all levels:
+        n_tiles = (4 ** (levels + 1) - 1) // 3
+
+        # Default all tiles to empty:
+        self._filedata = [pack("<II", 0, 0)] * n_tiles
+
+        # Reserve space for the file data (and the 8 header bytes):
+        self._next_offset = (n_tiles + 1) * 8
+        stream.seek(self._next_offset)
+
+    def close(self):
+        """Close the writer.
+
+        This writes out the index of tile data and closes
+        the underlying stream, making this object unusable
+        for future I/O."""
+
+        if self._stream is None:
+            return  # should only happen if we already close()d, which is OK
+
+        self._stream.seek(8)
+
+        for entry in self._filedata:
+            self._stream.write(entry)
+
+        self._stream.close()
+        self._stream = None
+
+    def __del__(self):
+        self.close()
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *_exc):
+        self.close()
+        return False
+
+    def append_stream(self, level: int, x: int, y: int, stream: BinaryIO):
+        """Append a tile to the plate file, getting data from a file-like
+        object.
+
+        Parameters
+        ----------
+        level : int
+            The level of the tile to write
+        x : int
+            The X position of the tile to write
+        y : int
+            The Y position of the tile to write
+        stream : readable, bytes-based file-like object
+            The source of tile data.
+
+        Returns
+        -------
+        length : int
+            The number of bytes read from *stream* and written to the plate
+            file.
+
+        Notes
+        -----
+        This method reads *stream* until end of file, but does not take
+        responsibility for closing it."""
+
+        if self._stream is None:
+            raise Exception("cannot write a closed V1PlateWriter")
+
+        if level < 0 or level > self._levels:
+            raise ValueError(f"invalid `level` {level}")
+
+        n = 2**level
+
+        if x < 0 or x >= n or y < 0 or y >= n:
+            raise ValueError(f"invalid tile position L{level}X{x}Y{y}")
+
+        if self._next_offset >= 4294967296:  # that's 2**32
+            raise Exception(
+                "cannot append to V1PlateWriter: 4-gibibyte size limit exceeded"
+            )
+
+        # This is basically `shutil.copyfileobj()`, but that doesn't tell
+        # us the total length written.
+
+        length = 0
+
+        while True:
+            b = stream.read(65536)
+            if not b:
+                break
+
+            self._stream.write(b)
+            length += len(b)
+
+        if length >= 4294967296:  # that's 2**32
+            raise Exception(
+                "error appending to V1PlateWriter: 4-gibibyte file size limit exceeded"
+            )
+
+        # Now we can add the filedata entry.
+        #
+        # This is the total number of tiles in all levels from 0 to `level - 1`:
+        index = (4**level - 1) // 3
+
+        # The offset of this tile within the level:
+        index += n * y + x
+
+        self._filedata[index] = pack("<II", self._next_offset, length)
+        self._next_offset += length
+        return length
+
+    def append_bytes(self, level: int, x: int, y: int, data: bytes):
+        """Append a tile to the plate file, getting data from a bytes buffer.
+
+        Parameters
+        ----------
+        level : int
+            The level of the tile to write
+        x : int
+            The X position of the tile to write
+        y : int
+            The Y position of the tile to write
+        data : bytes
+            The tile data.
+
+        Returns
+        -------
+        length : int
+            The number of bytes written to the plate file, which is the length
+            of *data*."""
+        return self.append_stream(level, x, y, BytesIO(data))
diff --git a/wwt_data_formats/tests/test_plate.py b/wwt_data_formats/tests/test_plate.py
new file mode 100644
index 0000000..6a27f54
--- /dev/null
+++ b/wwt_data_formats/tests/test_plate.py
@@ -0,0 +1,50 @@
+# -*- mode: python; coding: utf-8 -*-
+# Copyright 2022 the .NET Foundation
+# Licensed under the MIT License.
+
+import os
+import pytest
+
+from .. import plate
+from . import work_in_tempdir
+
+
+def test_basic_v1_plate(work_in_tempdir):
+    with plate.V1PlateWriter(open("test.plate", "wb"), 1) as pw:
+        assert pw.append_bytes(1, 1, 1, b"111") == 3
+        assert pw.append_bytes(1, 1, 0, b"110") == 3
+        assert pw.append_bytes(1, 0, 1, b"101") == 3
+        assert pw.append_bytes(1, 0, 0, b"100") == 3
+        assert pw.append_bytes(0, 0, 0, b"000") == 3
+
+        pw.close()
+
+        with pytest.raises(Exception):
+            pw.append_bytes(0, 0, 0, b"000")
+
+    # 8 bytes * (5 tiles + 1 header) + 3 bytes * 5 tiles =>
+    assert os.stat("test.plate").st_size == 63
+
+    with plate.V1PlateReader(open("test.plate", "rb")) as pr:
+        assert pr.read_tile(1, 1, 1) == b"111"
+        assert pr.read_tile(1, 1, 0) == b"110"
+        assert pr.read_tile(1, 0, 1) == b"101"
+        assert pr.read_tile(1, 0, 0) == b"100"
+        assert pr.read_tile(0, 0, 0) == b"000"
+
+        with pytest.raises(ValueError):
+            pr.read_tile(-1, 0, 0)
+
+        with pytest.raises(ValueError):
+            pr.read_tile(2, 0, 0)
+
+        with pytest.raises(ValueError):
+            pr.read_tile(1, -1, 0)
+
+        with pytest.raises(ValueError):
+            pr.read_tile(1, 0, 2)
+
+        pr.close()
+
+        with pytest.raises(Exception):
+            pr.read_tile(1, 0, 0)

From 531df450adc745a13335a04510cf76dd77744d17 Mon Sep 17 00:00:00 2001
From: Peter Williams <peter@newton.cx>
Date: Wed, 14 Sep 2022 15:54:36 -0400
Subject: [PATCH 4/4] ci/azure-build-and-test.yml: update macos image version

---
 ci/azure-build-and-test.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/ci/azure-build-and-test.yml b/ci/azure-build-and-test.yml
index 1fdfc8a..965ec87 100644
--- a/ci/azure-build-and-test.yml
+++ b/ci/azure-build-and-test.yml
@@ -16,12 +16,12 @@ parameters:
       PYTHON_SERIES: "3.8"
 
   - name: macos_37
-    vmImage: macos-10.15
+    vmImage: macos-11
     vars:
       PYTHON_SERIES: "3.7"
 
   - name: macos_38
-    vmImage: macos-10.15
+    vmImage: macos-11
     vars:
       PYTHON_SERIES: "3.8"