From 1fc3508338859b492e98ce63c8d9740e57280f85 Mon Sep 17 00:00:00 2001 From: William Ayd Date: Wed, 15 Jan 2020 15:06:01 -0800 Subject: [PATCH] Release notes (#80) --- doc/source/caveats.rst | 14 +++----------- doc/source/changelog.rst | 21 +++++++++++++++++++++ pantab/__init__.py | 2 +- pantab/newsfragments/20.feature | 5 ----- pantab/newsfragments/52.bugfix | 1 - setup.py | 2 +- 6 files changed, 26 insertions(+), 19 deletions(-) delete mode 100644 pantab/newsfragments/20.feature delete mode 100644 pantab/newsfragments/52.bugfix diff --git a/doc/source/caveats.rst b/doc/source/caveats.rst index 2860b034..85c093da 100644 --- a/doc/source/caveats.rst +++ b/doc/source/caveats.rst @@ -1,5 +1,5 @@ -Usage Limitations -================= +Usage Notes +=========== Type Mapping ------------ @@ -43,20 +43,12 @@ pantab maps the following dtypes from pandas to the equivalent column type in Ta Any dtype not explicitly listed in the above table will raise a ValueError if trying to write out data. .. versionadded:: 1.0.0 - If using pandas 1.0 and above, text data will be read back into a ``"string"`` dtype rather than an ``object`` dtype. + If using pandas 1.0 and above, text data will be read back into a ``string`` dtype rather than an ``object`` dtype. .. note:: Most objects can maintain their type when "round-tripping" to/from Hyper extracts, with the exception of the float32 object as only DOUBLE is available for floating point storage in Hyper. After pandas 1.0 / pantab 1.0, object dtypes written will be read back in as string. -Missing Value Handling ----------------------- -In the scientific Python community missing values are often notated as ``np.nan``, which is a float value implementing the `IEEE 754 `_ standard for missing value handling. Because NumPy stores homogenous arrays, there has been a long standing issue in pandas where integers could not store missing records, as noted in the `pandas documentation on missing values `_. - -Complicating factors is the presence of the ``None`` sentinel in Python, which can often times intuitively be represented as a missing value. Unfortunately ``None`` is a Python level construct, and most of the performance of the data stack in Python is implemented in lower-level C routines that cannot take advantage of this value. The Hyper API though uses this as its missing value indicator. - -Because of the various missing value indicators, serializing to/from a Hyper extract via pantab cannot be lossless. On write, ``None`` and ``np.nan`` values will both be written in as missing data to the Hyper extract. On read, missing values will be read back as ``np.nan``. - Index / Column Handling ----------------------- A pandas DataFrame always comes with an ``Index`` used to slice / access rows. No such concept exists in Tableau, so this will be implicitly dropped when writing Hyper extracts. diff --git a/doc/source/changelog.rst b/doc/source/changelog.rst index 46e6eeb1..bd37907d 100644 --- a/doc/source/changelog.rst +++ b/doc/source/changelog.rst @@ -1,6 +1,27 @@ Changelog ^^^^^^^^^ +Pantab 1.0.0 (2020-01-15) +========================= + +Special thanks to `chillerno1 `_ for contributing to this release. + +Features +-------- + +- pantab now supports reading/writing pandas 1.0 dtypes, namely the ``boolean`` and ``string`` dtypes. (#20) + + .. important:: + + TEXT data read from a Hyper extract will be stored in a ``string`` dtype when using pandas 1.0 or greater in combination with pantab 1.0 or greater. Older versions of either tool will read the data back into a ``object`` dtype. + + +Bugfixes +-------- + +- Fixed potential segfault on systems where not all addresses can be expressed in an unsigned long long. (#52) + + Pantab 0.2.3 (2020-01-02) ========================= diff --git a/pantab/__init__.py b/pantab/__init__.py index b7066f2c..f2d0f404 100644 --- a/pantab/__init__.py +++ b/pantab/__init__.py @@ -1,4 +1,4 @@ -__version__ = "0.2.3" +__version__ = "1.0.0" from ._reader import frame_from_hyper, frames_from_hyper diff --git a/pantab/newsfragments/20.feature b/pantab/newsfragments/20.feature deleted file mode 100644 index f74caf63..00000000 --- a/pantab/newsfragments/20.feature +++ /dev/null @@ -1,5 +0,0 @@ -pantab now supports reading/writing pandas 1.0 dtypes, namely the ``boolean`` and ``string`` dtypes. - -.. important:: - - TEXT data read from a Hyper extract will be stored in a ``"string"`` dtype when using pandas 1.0 or greater in combination with pantab 1.0 or greater. Older versions of either tool will read the data back into a ``object`` dtype. \ No newline at end of file diff --git a/pantab/newsfragments/52.bugfix b/pantab/newsfragments/52.bugfix deleted file mode 100644 index 0a8622de..00000000 --- a/pantab/newsfragments/52.bugfix +++ /dev/null @@ -1 +0,0 @@ -Fixed potential segfault on systems where not all addresses can be expressed in an unsigned long long. \ No newline at end of file diff --git a/setup.py b/setup.py index e7dab000..319bb06c 100644 --- a/setup.py +++ b/setup.py @@ -63,7 +63,7 @@ setup( name="pantab", - version="0.2.3", + version="1.0.0", description="Converts pandas DataFrames into Tableau Hyper Extracts", long_description=long_description, long_description_content_type="text/markdown",