Internal Design

Dask DataFrames coordinate many Pandas DataFrames/Series arranged along an index. We define a Dask DataFrame object with the following components:

  • A Dask graph with a special set of keys designating partitions, such as ('x', 0), ('x', 1), ...
  • A name to identify which keys in the Dask graph refer to this DataFrame, such as 'x'
  • An empty Pandas object containing appropriate metadata (e.g. column names, dtypes, etc.)
  • A sequence of partition boundaries along the index called divisions


Many DataFrame operations rely on knowing the name and dtype of columns. To keep track of this information, all Dask DataFrame objects have a _meta attribute which contains an empty Pandas object with the same dtypes and names. For example:

>>> df = pd.DataFrame({'a': [1, 2, 3], 'b': ['x', 'y', 'z']})
>>> ddf = dd.from_pandas(df, npartitions=2)
>>> ddf._meta
Empty DataFrame
Columns: [a, b]
Index: []
>>> ddf._meta.dtypes
a     int64
b    object
dtype: object

Internally, Dask DataFrame does its best to propagate this information through all operations, so most of the time a user shouldn’t have to worry about this. Usually this is done by evaluating the operation on a small sample of fake data, which can be found on the _meta_nonempty attribute:

>>> ddf._meta_nonempty
   a    b
0  1  foo
1  1  foo

Sometimes this operation may fail in user defined functions (e.g. when using DataFrame.apply), or may be prohibitively expensive. For these cases, many functions support an optional meta keyword, which allows specifying the metadata directly, avoiding the inference step. For convenience, this supports several options:

  1. A Pandas object with appropriate dtypes and names. If not empty, an empty slice will be taken:
>>> ddf.map_partitions(foo, meta=pd.DataFrame({'a': [1], 'b': [2]}))
  1. A description of the appropriate names and dtypes. This can take several forms:

    • A dict of {name: dtype} or an iterable of (name, dtype) specifies a DataFrame
    • A tuple of (name, dtype) specifies a series
    • A dtype object or string (e.g. 'f8') specifies a scalar

This keyword is available on all functions/methods that take user provided callables (e.g. DataFrame.map_partitions, DataFrame.apply, etc…), as well as many creation functions (e.g. dd.from_delayed).


Dask DataFrame divides categorical data into two types:

  • Known categoricals have the categories known statically (on the _meta attribute). Each partition must have the same categories as found on the _meta attribute
  • Unknown categoricals don’t know the categories statically, and may have different categories in each partition. Internally, unknown categoricals are indicated by the presence of dd.utils.UNKNOWN_CATEGORIES in the categories on the _meta attribute. Since most DataFrame operations propagate the categories, the known/unknown status should propagate through operations (similar to how NaN propagates)

For metadata specified as a description (option 2 above), unknown categoricals are created.

Certain operations are only available for known categoricals. For example, would only work if df.col has known categories, since the categorical mapping is only known statically on the metadata of known categoricals.

The known/unknown status for a categorical column can be found using the known property on the categorical accessor:


Additionally, an unknown categorical can be converted to known using .cat.as_known(). If you have multiple categorical columns in a DataFrame, you may instead want to use df.categorize(columns=...), which will convert all specified columns to known categoricals. Since getting the categories requires a full scan of the data, using df.categorize() is more efficient than calling .cat.as_known() for each column (which would result in multiple scans):

>>> col_known =  # use for single column
>>> ddf_known = ddf.categorize()        # use for multiple columns

To convert a known categorical to an unknown categorical, there is also the .cat.as_unknown() method. This requires no computation as it’s just a change in the metadata.

Non-categorical columns can be converted to categoricals in a few different ways:

# astype operates lazily, and results in unknown categoricals
ddf = ddf.astype({'mycol': 'category', ...})
# or
ddf['mycol'] = ddf.mycol.astype('category')

# categorize requires computation, and results in known categoricals
ddf = ddf.categorize(columns=['mycol', ...])

Additionally, with Pandas 0.19.2 and up, dd.read_csv and dd.read_table can read data directly into unknown categorical columns by specifying a column dtype as 'category':

>>> ddf = dd.read_csv(..., dtype={col_name: 'category'})

Moreover, with Pandas 0.21.0 and up, dd.read_csv and dd.read_table can read data directly into known categoricals by specifying instances of pd.api.types.CategoricalDtype:

>>> dtype = {'col': pd.api.types.CategoricalDtype(['a', 'b', 'c'])}
>>> ddf = dd.read_csv(..., dtype=dtype)


Internally, a Dask DataFrame is split into many partitions, where each partition is one Pandas DataFrame. These DataFrames are split vertically along the index. When our index is sorted and we know the values of the divisions of our partitions, then we can be clever and efficient with expensive algorithms (e.g. groupby’s, joins, etc…).

For example, if we have a time-series index, then our partitions might be divided by month: all of January will live in one partition while all of February will live in the next. In these cases, operations like loc, groupby, and join/merge along the index can be much more efficient than would otherwise be possible in parallel. You can view the number of partitions and divisions of your DataFrame with the following fields:

>>> df.npartitions
>>> df.divisions
['2015-01-01', '2015-02-01', '2015-03-01', '2015-04-01', '2015-04-31']

Divisions includes the minimum value of every partition’s index and the maximum value of the last partition’s index. In the example above, if the user searches for a specific datetime range, then we know which partitions we need to inspect and which we can drop:

>>> df.loc['2015-01-20': '2015-02-10']  # Must inspect first two partitions

Often we do not have such information about our partitions. When reading CSV files, for example, we do not know, without extra user input, how the data is divided. In this case .divisions will be all None:

>>> df.divisions
[None, None, None, None, None]

In these cases, any operation that requires a cleanly partitioned DataFrame with known divisions will have to perform a sort. This can generally achieved by calling df.set_index(...).