Home

Welcome to the fletcher wiki!

We mainly use this space to collect links to related projects, e.g. upstream issues that we would like to see fixed in general.

Apache Arrow issues

• Add to_pandas to ChunkedArray 0.10.0
• Make dictionary_encode and unique accesible on Column / ChunkedArray 0.10.0
• Implement getitem / slicing on Buffer 0.10.0
• Add index() method to find first occurence of Python scalar
• numpy.asarray should trigger to_pandas on Array/ChunkedArray 0.10.0
• [C++] Implement take kernel functions 0.16.0
• Add pandas-like take method to Array/Column/ChunkedArray 0.14.0
• Add nicer repr for Column / ChunkedArray 0.10.0
• Variable step size slicing for arrays
• pa.array from np.datetime[D] and type=pa.date64 produces invalid results 0.10.0
• Inconsistent handling of np.nan 0.10.0
• Casting float NaN to int should raise an error on safe cast 0.11.0

Pandas issues

• Series[EA].astype always casts intermediate to NumPy array 0.24.0
• Direct delegation of Series methods to ExtensionArrays
• pd.concat of Series with int64 column and Serias with int64-ExtensionArray yields int64
• Accept constant memoryviews in HashTable.lookup 0.24.0
• _reduce needs to be defined on non-NumPy ExtensionArrays / add reduce op (and groupby reduce) support to EA 0.24.0
• BUG: pandas.core.ops.dispatch_to_extension_op fails with UnboundLocalError 0.24.0
• BUG: ExtensionArray formatting of datetime-like forces nanosecond precision
• BUG: concat_{categorical, compat} leads to erroneous result on non-ns datetime-EA