Released on March 5, 2021


This is the first release with support for Python 3.9 and the last release with support for Python 3.6


Released on February 5, 2021


Released on January 22, 2021


Released on January 15, 2021


Released on December 10, 2020


  • Switched to CalVer for versioning scheme.
  • Introduced new APIs for HighLevelGraph to enable sending high-level representations of task graphs to the distributed scheduler.
  • Introduced new HighLevelGraph layer objects including BasicLayer, Blockwise, BlockwiseIO, ShuffleLayer, and more.
  • Added support for applying custom Layer-level annotations like priority, retries, etc. with the dask.annotations context manager.
  • Updated minimum supported version of pandas to 0.25.0 and NumPy to 1.15.1.
  • Support for the pyarrow.dataset API to read_parquet.
  • Several fixes to Dask Array’s SVD.

All changes

2.30.0 / 2020-10-06

2.29.0 / 2020-10-02

2.28.0 / 2020-09-25

2.27.0 / 2020-09-18

2.26.0 / 2020-09-11

2.25.0 / 2020-08-28

2.24.0 / 2020-08-22

2.23.0 / 2020-08-14

2.22.0 / 2020-07-31

2.21.0 / 2020-07-17

2.20.0 / 2020-07-02

2.19.0 / 2020-06-19

2.18.1 / 2020-06-09

2.18.0 / 2020-06-05

2.17.2 / 2020-05-28

2.17.1 / 2020-05-28

2.17.0 / 2020-05-26

2.16.0 / 2020-05-08

2.15.0 / 2020-04-24

2.14.0 / 2020-04-03

2.13.0 / 2020-03-25

2.12.0 / 2020-03-06

2.11.0 / 2020-02-19

2.10.1 / 2020-01-30

2.10.0 / 2020-01-28

2.9.2 / 2020-01-16

2.9.1 / 2019-12-27

2.9.0 / 2019-12-06

2.8.1 / 2019-11-22

2.8.0 / 2019-11-14

2.7.0 / 2019-11-08

This release drops support for Python 3.5

2.6.0 / 2019-10-15

2.5.2 / 2019-10-04

2.5.0 / 2019-09-27

2.4.0 / 2019-09-13

2.3.0 / 2019-08-16

2.2.0 / 2019-08-01

2.1.0 / 2019-07-08

2.0.0 / 2019-06-25

1.2.2 / 2019-05-08

1.2.1 / 2019-04-29

1.2.0 / 2019-04-12

1.1.5 / 2019-03-29

1.1.4 / 2019-03-08

1.1.3 / 2019-03-01

1.1.2 / 2019-02-25

1.1.1 / 2019-01-31

  • Work around psutil 5.5.0 not allowing pickling Process objects Janne Vuorela

1.1.0 / 2019-01-18

1.0.0 / 2018-11-28

0.20.2 / 2018-11-15

0.20.1 / 2018-11-09

0.20.0 / 2018-10-26

0.19.4 / 2018-10-09

0.19.3 / 2018-10-05

0.19.2 / 2018-09-17

0.19.1 / 2018-09-06

0.19.0 / 2018-08-29

0.18.2 / 2018-07-23

0.18.1 / 2018-06-22

0.18.0 / 2018-06-14

0.17.5 / 2018-05-16

0.17.4 / 2018-05-03

0.17.3 / 2018-05-02

  • Support traversing collections in persist, visualize, and optimize (GH#3410) Jim Crist
  • Add schedule= keyword to compute and persist. This replaces common use of the get= keyword (GH#3448) Matthew Rocklin

0.17.2 / 2018-03-21

0.17.1 / 2018-02-22

0.17.0 / 2018-02-09

  • Document bag.map_paritions function may receive either a list or generator. (GH#3150) Nir

0.16.1 / 2018-01-09

0.16.0 / 2017-11-17

This is a major release. It includes breaking changes, new protocols, and a large number of bug fixes.

0.15.4 / 2017-10-06

  • da.random.choice now works with array arguments (GH#2781)
  • Support indexing in arrays with (fixes regression) (GH#2719)
  • Handle zero dimension with rechunking (GH#2747)
  • Support -1 as an alias for “size of the dimension” in chunks (GH#2749)
  • Call mkdir in array.to_npy_stack (GH#2709)
  • Added the .str accessor to Categoricals with string categories (GH#2743)
  • Support int96 (spark) datetimes in parquet writer (GH#2711)
  • Pass on file scheme to fastparquet (GH#2714)
  • Support Pandas 0.21 (GH#2737)
  • Add tree reduction support for foldby (GH#2710)
  • Drop s3fs from pip install dask[complete] (GH#2750)

0.15.3 / 2017-09-24

  • Add masked arrays (GH#2301)
  • Add *_like array creation functions (GH#2640)
  • Indexing with unsigned integer array (GH#2647)
  • Improved slicing with boolean arrays of different dimensions (GH#2658)
  • Support literals in top and atop (GH#2661)
  • Optional axis argument in cumulative functions (GH#2664)
  • Improve tests on scalars with assert_eq (GH#2681)
  • Fix norm keepdims (GH#2683)
  • Add ptp (GH#2691)
  • Add apply_along_axis (GH#2690) and apply_over_axes (GH#2702)
  • Added Series.str[index] (GH#2634)
  • Allow the groupby by param to handle columns and index levels (GH#2636)
  • DataFrame.to_csv and Bag.to_textfiles now return the filenames to
    which they have written (GH#2655)
  • Fix combination of partition_on and append in to_parquet (GH#2645)
  • Fix for parquet file schemes (GH#2667)
  • Repartition works with mixed categoricals (GH#2676)
  • python test now runs tests (GH#2641)
  • Added new cheatsheet (GH#2649)
  • Remove resize tool in Bokeh plots (GH#2688)

0.15.2 / 2017-08-25

  • Remove spurious keys from map_overlap graph (GH#2520)
  • where works with non-bool condition and scalar values (GH#2543) (GH#2549)
  • Improve compress (GH#2541) (GH#2545) (GH#2555)
  • Add argwhere, _nonzero, and where(cond) (GH#2539)
  • Generalize vindex in dask.array to handle multi-dimensional indices (GH#2573)
  • Add choose method (GH#2584)
  • Split code into reorganized files (GH#2595)
  • Add linalg.norm (GH#2597)
  • Add diff, ediff1d (GH#2607), (GH#2609)
  • Improve dtype inference and reflection (GH#2571)
  • Remove deprecated Bag behaviors (GH#2525)
  • Remove bare except: blocks everywhere (GH#2590)

0.15.1 / 2017-07-08

  • Add storage_options to to_textfiles and to_csv (GH#2466)
  • Rechunk and simplify rfftfreq (GH#2473), (GH#2475)
  • Better support ndarray subclasses (GH#2486)
  • Import star in dask.distributed (GH#2503)
  • Threadsafe cache handling with tokenization (GH#2511)

0.15.0 / 2017-06-09

  • Fix bug where reductions on bags with no partitions would fail (GH#2324)
  • Add broadcasting and variadic top-level function. Also remove auto-expansion of tuples as map arguments (GH#2339)
  • Rename Bag.concat to Bag.flatten (GH#2402)
  • Move dask.async module to dask.local (GH#2318)
  • Support callbacks with nested scheduler calls (GH#2397)
  • Support pathlib.Path objects as uris (GH#2310)

0.14.3 / 2017-05-05

  • Pandas 0.20.0 support

0.14.2 / 2017-05-03

  • Add da.indices (GH#2268), da.tile (GH#2153), da.roll (GH#2135)
  • Simultaneously support drop_axis and new_axis in da.map_blocks (GH#2264)
  • Rechunk and concatenate work with unknown chunksizes (GH#2235) and (GH#2251)
  • Support non-numpy container arrays, notably sparse arrays (GH#2234)
  • Tensordot contracts over multiple axes (GH#2186)
  • Allow delayed targets in (GH#2181)
  • Support interactions against lists and tuples (GH#2148)
  • Constructor plugins for debugging (GH#2142)
  • Multi-dimensional FFTs (single chunk) (GH#2116)
  • to_dataframe enforces consistent types (GH#2199)
  • Set_index always fully sorts the index (GH#2290)
  • Support compatibility with pandas 0.20.0 (GH#2249), (GH#2248), and (GH#2246)
  • Support Arrow Parquet reader (GH#2223)
  • Time-based rolling windows (GH#2198)
  • Repartition can now create more partitions, not just less (GH#2168)
  • Always use absolute paths when on POSIX file system (GH#2263)
  • Support user provided graph optimizations (GH#2219)
  • Refactor path handling (GH#2207)
  • Improve fusion performance (GH#2129), (GH#2131), and (GH#2112)

0.14.1 / 2017-03-22

  • Micro-optimize optimizations (GH#2058)
  • Change slicing optimizations to avoid fusing raw numpy arrays (GH#2075) (GH#2080)
  • Dask.array operations now work on numpy arrays (GH#2079)
  • Reshape now works in a much broader set of cases (GH#2089)
  • Support deepcopy python protocol (GH#2090)
  • Allow user-provided FFT implementations in da.fft (GH#2093)
  • Fix to_parquet with empty partitions (GH#2020)
  • Optional npartitions='auto' mode in set_index (GH#2025)
  • Optimize shuffle performance (GH#2032)
  • Support efficient repartitioning along time windows like repartition(freq='12h') (GH#2059)
  • Improve speed of categorize (GH#2010)
  • Support single-row dataframe arithmetic (GH#2085)
  • Automatically avoid shuffle when setting index with a sorted column (GH#2091)
  • Improve handling of integer-na handling in read_csv (GH#2098)
  • Repeated attribute access on delayed objects uses the same key (GH#2084)
  • Improve naming of nodes in dot visuals to avoid generic apply (GH#2070)
  • Ensure that worker processes have different random seeds (GH#2094)

0.14.0 / 2017-02-24

  • Repartition can now increase number of partitions (GH#1934)
  • Fix bugs in some reductions with empty partitions (GH#1939), (GH#1950), (GH#1953)
  • Support non-uniform categoricals (GH#1877), (GH#1930)
  • Groupby cumulative reductions (GH#1909)
  • DataFrame.loc indexing now supports lists (GH#1913)
  • Improve multi-level groupbys (GH#1914)
  • Improved HTML and string repr for DataFrames (GH#1637)
  • Parquet append (GH#1940)
  • Add dd.demo.daily_stock function for teaching (GH#1992)
  • Add traverse= keyword to delayed to optionally avoid traversing nested data structures (GH#1899)
  • Support Futures in from_delayed functions (GH#1961)
  • Improve serialization of decorated delayed functions (GH#1969)
  • Improve windows path parsing in corner cases (GH#1910)
  • Rename tasks when fusing (GH#1919)
  • Add top level persist function (GH#1927)
  • Propagate errors= keyword in byte handling (GH#1954)
  • Dask.compute traverses Python collections (GH#1975)
  • Structural sharing between graphs in dask.array and dask.delayed (GH#1985)

0.13.0 / 2017-01-02

  • Mandatory dtypes on dask.array. All operations maintain dtype information and UDF functions like map_blocks now require a dtype= keyword if it can not be inferred. (GH#1755)
  • Support arrays without known shapes, such as arises when slicing arrays with arrays or converting dataframes to arrays (GH#1838)
  • Support mutation by setting one array with another (GH#1840)
  • Tree reductions for covariance and correlations. (GH#1758)
  • Add SerializableLock for better use with distributed scheduling (GH#1766)
  • Improved atop support (GH#1800)
  • Rechunk optimization (GH#1737), (GH#1827)
  • Avoid wrong results when recomputing the same groupby twice (GH#1867)
  • Changed behaviour for delayed(nout=0) and delayed(nout=1): delayed(nout=1) does not default to out=None anymore, and delayed(nout=0) is also enabled. I.e. functions with return tuples of length 1 or 0 can be handled correctly. This is especially handy, if functions with a variable amount of outputs are wrapped by delayed. E.g. a trivial example: delayed(lambda *args: args, nout=len(vals))(*vals)

0.12.0 / 2016-11-03

  • Return a series when functions given to dataframe.map_partitions return scalars (GH#1515)
  • Fix type size inference for series (GH#1513)
  • dataframe.DataFrame.categorize no longer includes missing values in the categories. This is for compatibility with a pandas change (GH#1565)
  • Fix head parser error in dataframe.read_csv when some lines have quotes (GH#1495)
  • Add dataframe.reduction and series.reduction methods to apply generic row-wise reduction to dataframes and series (GH#1483)
  • Add dataframe.select_dtypes, which mirrors the pandas method (GH#1556)
  • dataframe.read_hdf now supports reading Series (GH#1564)
  • Support Pandas 0.19.0 (GH#1540)
  • Implement select_dtypes (GH#1556)
  • String accessor works with indexes (GH#1561)
  • Add pipe method to dask.dataframe (GH#1567)
  • Add indicator keyword to merge (GH#1575)
  • Support Series in read_hdf (GH#1575)
  • Support Categories with missing values (GH#1578)
  • Support inplace operators like df.x += 1 (GH#1585)
  • Str accessor passes through args and kwargs (GH#1621)
  • Improved groupby support for single-machine multiprocessing scheduler (GH#1625)
  • Tree reductions (GH#1663)
  • Pivot tables (GH#1665)
  • Add clip (GH#1667), align (GH#1668), combine_first (GH#1725), and any/all (GH#1724)
  • Improved handling of divisions on dask-pandas merges (GH#1666)
  • Add groupby.aggregate method (GH#1678)
  • Add dd.read_table function (GH#1682)
  • Improve support for multi-level columns (GH#1697) (GH#1712)
  • Support 2d indexing in loc (GH#1726)
  • Extend resample to include DataFrames (GH#1741)
  • Support dask.array ufuncs on dask.dataframe objects (GH#1669)
  • Add information about how dask.array chunks argument work (GH#1504)
  • Fix field access with non-scalar fields in dask.array (GH#1484)
  • Add concatenate= keyword to atop to concatenate chunks of contracted dimensions
  • Optimized slicing performance (GH#1539) (GH#1731)
  • Extend atop with a concatenate= (GH#1609) new_axes= (GH#1612) and adjust_chunks= (GH#1716) keywords
  • Add clip (GH#1610) swapaxes (GH#1611) round (GH#1708) repeat
  • Automatically align chunks in atop-backed operations (GH#1644)
  • Cull dask.arrays on slicing (GH#1709)
  • Fix issue with callables in bag.from_sequence being interpreted as tasks (GH#1491)
  • Avoid non-lazy memory use in reductions (GH#1747)
  • Added changelog (GH#1526)
  • Create new threadpool when operating from thread (GH#1487)
  • Unify example documentation pages into one (GH#1520)
  • Add versioneer for git-commit based versions (GH#1569)
  • Pass through node_attr and edge_attr keywords in dot visualization (GH#1614)
  • Add continuous testing for Windows with Appveyor (GH#1648)
  • Remove use of multiprocessing.Manager (GH#1653)
  • Add global optimizations keyword to compute (GH#1675)
  • Micro-optimize get_dependencies (GH#1722)

0.11.0 / 2016-08-24

DataFrames now enforce knowing full metadata (columns, dtypes) everywhere. Previously we would operate in an ambiguous state when functions lost dtype information (such as apply). Now all dataframes always know their dtypes and raise errors asking for information if they are unable to infer (which they usually can). Some internal attributes like _pd and _pd_nonempty have been moved.

The internals of the distributed scheduler have been refactored to transition tasks between explicit states. This improves resilience, reasoning about scheduling, plugin operation, and logging. It also makes the scheduler code easier to understand for newcomers.

  • The distributed.s3 and distributed.hdfs namespaces are gone. Use protocols in normal methods like read_text('s3://...' instead.
  • Dask.array.reshape now errs in some cases where previously it would have create a very large number of tasks

0.10.2 / 2016-07-27

  • More Dataframe shuffles now work in distributed settings, ranging from setting-index to hash joins, to sorted joins and groupbys.
  • Dask passes the full test suite when run when under in Python’s optimized-OO mode.
  • On-disk shuffles were found to produce wrong results in some highly-concurrent situations, especially on Windows. This has been resolved by a fix to the partd library.
  • Fixed a growth of open file descriptors that occurred under large data communications
  • Support ports in the --bokeh-whitelist option ot dask-scheduler to better routing of web interface messages behind non-trivial network settings
  • Some improvements to resilience to worker failure (though other known failures persist)
  • You can now start an IPython kernel on any worker for improved debugging and analysis
  • Improvements to dask.dataframe.read_hdf, especially when reading from multiple files and docs

0.10.0 / 2016-06-13

  • This version drops support for Python 2.6
  • Conda packages are built and served from conda-forge
  • The dask.distributed executables have been renamed from dfoo to dask-foo. For example dscheduler is renamed to dask-scheduler
  • Both Bag and DataFrame include a preliminary distributed shuffle.
  • Add task-based shuffle for distributed groupbys
  • Add accumulate for cumulative reductions
  • Add a task-based shuffle suitable for distributed joins, groupby-applys, and set_index operations. The single-machine shuffle remains untouched (and much more efficient.)
  • Add support for new Pandas rolling API with improved communication performance on distributed systems.
  • Add groupby.std/var
  • Pass through S3/HDFS storage options in read_csv
  • Improve categorical partitioning
  • Add eval, info, isnull, notnull for dataframes
  • Rename executables like dscheduler to dask-scheduler
  • Improve scheduler performance in the many-fast-tasks case (important for shuffling)
  • Improve work stealing to be aware of expected function run-times and data sizes. The drastically increases the breadth of algorithms that can be efficiently run on the distributed scheduler without significant user expertise.
  • Support maximum buffer sizes in streaming queues
  • Improve Windows support when using the Bokeh diagnostic web interface
  • Support compression of very-large-bytestrings in protocol
  • Support clean cancellation of submitted futures in Joblib interface
  • All dask-related projects (dask, distributed, s3fs, hdfs, partd) are now building conda packages on conda-forge.
  • Change credential handling in s3fs to only pass around delegated credentials if explicitly given secret/key. The default now is to rely on managed environments. This can be changed back by explicitly providing a keyword argument. Anonymous mode must be explicitly declared if desired.

0.9.0 / 2016-05-11

  • and dask.value have been renamed to dask.delayed
  • dask.bag.from_filenames has been renamed to dask.bag.read_text
  • All S3/HDFS data ingest functions like db.from_s3 or distributed.s3.read_csv have been moved into the plain read_text, read_csv functions, which now support protocols, like dd.read_csv('s3://bucket/keys*.csv')
  • Add support for scipy.LinearOperator
  • Improve optional locking to on-disk data structures
  • Change rechunk to expose the intermediate chunks
  • Rename from_filenames to read_text
  • Remove from_s3 in favor of read_text('s3://...')
  • Fixed numerical stability issue for correlation and covariance
  • Allow no-hash from_pandas for speedy round-trips to and from-pandas objects
  • Generally reengineered read_csv to be more in line with Pandas behavior
  • Support fast set_index operations for sorted columns
  • Rename do/value to delayed
  • Rename to/from_imperative to to/from_delayed
  • Move s3 and hdfs functionality into the dask repository
  • Adaptively oversubscribe workers for very fast tasks
  • Improve PyPy support
  • Improve work stealing for unbalanced workers
  • Scatter data efficiently with tree-scatters
  • Add lzma/xz compression support
  • Raise a warning when trying to split unsplittable compression types, like gzip or bz2
  • Improve hashing for single-machine shuffle operations
  • Add new callback method for start state
  • General performance tuning

0.8.1 / 2016-03-11

  • Bugfix for range slicing that could periodically lead to incorrect results.
  • Improved support and resiliency of arg reductions (argmin, argmax, etc.)
  • Add zip function
  • Add corr and cov functions
  • Add melt function
  • Bugfixes for io to bcolz and hdf5

0.8.0 / 2016-02-20

  • Changed default array reduction split from 32 to 4
  • Linear algebra, tril, triu, LU, inv, cholesky, solve, solve_triangular, eye, lstsq, diag, corrcoef.
  • Add tree reductions
  • Add range function
  • drop from_hdfs function (better functionality now exists in hdfs3 and distributed projects)
  • Refactor dask.dataframe to include a full empty pandas dataframe as metadata. Drop the .columns attribute on Series
  • Add Series categorical accessor, series.nunique, drop the .columns attribute for series.
  • read_csv fixes (multi-column parse_dates, integer column names, etc. )
  • Internal changes to improve graph serialization
  • Documentation updates
  • Add from_imperative and to_imperative functions for all collections
  • Aesthetic changes to profiler plots
  • Moved the dask project to a new dask organization

0.7.6 / 2016-01-05

  • Improve thread safety
  • Tree reductions
  • Add view, compress, hstack, dstack, vstack methods
  • map_blocks can now remove and add dimensions
  • Improve thread safety
  • Extend sampling to include replacement options
  • Removed optimization passes that fused results.
  • Removed dask.distributed
  • Improved performance of blocked file reading
  • Serialization improvements
  • Test Python 3.5

0.7.4 / 2015-10-23

This was mostly a bugfix release. Some notable changes:

  • Fix minor bugs associated with the release of numpy 1.10 and pandas 0.17
  • Fixed a bug with random number generation that would cause repeated blocks due to the birthday paradox
  • Use locks in dask.dataframe.read_hdf by default to avoid concurrency issues
  • Change dask.get to point to dask.async.get_sync by default
  • Allow visualization functions to accept general graphviz graph options like rankdir=’LR’
  • Add reshape and ravel to dask.array
  • Support the creation of dask.arrays from dask.imperative objects

This release also includes a deprecation warning for dask.distributed, which will be removed in the next version.

Future development in distributed computing for dask is happening here: . General feedback on that project is most welcome from this community.

0.7.3 / 2015-09-25

  • A utility for profiling memory and cpu usage has been added to the dask.diagnostics module.

This release improves coverage of the pandas API. Among other things it includes nunique, nlargest, quantile. Fixes encoding issues with reading non-ascii csv files. Performance improvements and bug fixes with resample. More flexible read_hdf with globbing. And many more. Various bug fixes in dask.imperative and dask.bag.

0.7.0 / 2015-08-15

This release includes significant bugfixes and alignment with the Pandas API. This has resulted both from use and from recent involvement by Pandas core developers.

  • New operations: query, rolling operations, drop
  • Improved operations: quantiles, arithmetic on full dataframes, dropna, constructor logic, merge/join, elemwise operations, groupby aggregations
  • Fixed a bug in fold where with a null default argument
  • New operations: da.fft module, da.image.imread
  • The array and dataframe collections create graphs with deterministic keys. These tend to be longer (hash strings) but should be consistent between computations. This will be useful for caching in the future.
  • All collections (Array, Bag, DataFrame) inherit from common subclass

0.6.1 / 2015-07-23

  • Improved (though not yet sufficient) resiliency for dask.distributed when workers die
  • Improved writing to various formats, including to_hdf, to_castra, and to_csv
  • Improved creation of dask DataFrames from dask Arrays and Bags
  • Improved support for categoricals and various other methods
  • Various bug fixes
  • Histogram function
  • Added tie-breaking ordering of tasks within parallel workloads to better handle and clear intermediate results
  • Added the function for explicit construction of graphs with normal python code
  • Traded pydot for graphviz library for graph printing to support Python3
  • There is also a gitter chat room and a stackoverflow tag