Delay import (pandas-dev#17710)

closes pandas-dev#16764

Author: TomAugspurger

.travis.yml

@@ -121,6 +121,8 @@ script:
   - ci/script_single.sh
   - ci/script_multi.sh
   - ci/lint.sh
+  - echo "checking imports"
+  - source activate pandas && python ci/check_imports.py
   - echo "script done"
 
 after_success:

ci/check_imports.py

@@ -0,0 +1,36 @@
+"""
+Check that certain modules are not loaded by `import pandas`
+"""
+import sys
+
+blacklist = {
+    'bs4',
+    'html5lib',
+    'ipython',
+    'jinja2'
+    'lxml',
+    'matplotlib',
+    'numexpr',
+    'openpyxl',
+    'py',
+    'pytest',
+    's3fs',
+    'scipy',
+    'tables',
+    'xlrd',
+    'xlsxwriter',
+    'xlwt',
+}
+
+
+def main():
+    import pandas  # noqa
+
+    modules = set(x.split('.')[0] for x in sys.modules)
+    imported = modules & blacklist
+    if modules & blacklist:
+        sys.exit("Imported {}".format(imported))
+
+
+if __name__ == '__main__':
+    main()

doc/source/whatsnew/v0.21.0.txt

@@ -164,6 +164,7 @@ Other Enhancements
 - :func:`read_excel` raises ``ImportError`` with a better message if ``xlrd`` is not installed. (:issue:`17613`)
 - :func:`read_json` now accepts a ``chunksize`` parameter that can be used when ``lines=True``. If ``chunksize`` is passed, read_json now returns an iterator which reads in ``chunksize`` lines with each iteration. (:issue:`17048`)
 - :meth:`DataFrame.assign` will preserve the original order of ``**kwargs`` for Python 3.6+ users instead of sorting the column names
+- Improved the import time of pandas by about 2.25x  (:issue:`16764`)
 
 
 .. _whatsnew_0210.api_breaking:
@@ -559,6 +560,8 @@ Other API Changes
 - :func:`to_datetime` when passed a tz-aware ``origin=`` kwarg will now raise a more informative ``ValueError`` rather than a ``TypeError`` (:issue:`16842`)
 - Renamed non-functional ``index`` to ``index_col`` in :func:`read_stata` to improve API consistency (:issue:`16342`)
 - Bug in :func:`DataFrame.drop` caused boolean labels ``False`` and ``True`` to be treated as labels 0 and 1 respectively when dropping indices from a numeric index. This will now raise a ValueError (:issue:`16877`)
+- Pandas no longer registers matplotlib converters on import. The converters
+  will be registered and used when the first plot is draw (:issue:`17710`)
 
 .. _whatsnew_0210.deprecations:
 

pandas/core/computation/__init__.py

@@ -1,23 +0,0 @@
-
-import warnings
-from distutils.version import LooseVersion
-
-_NUMEXPR_INSTALLED = False
-_MIN_NUMEXPR_VERSION = "2.4.6"
-
-try:
-    import numexpr as ne
-    ver = ne.__version__
-    _NUMEXPR_INSTALLED = ver >= LooseVersion(_MIN_NUMEXPR_VERSION)
-
-    if not _NUMEXPR_INSTALLED:
-        warnings.warn(
-            "The installed version of numexpr {ver} is not supported "
-            "in pandas and will be not be used\nThe minimum supported "
-            "version is {min_ver}\n".format(
-                ver=ver, min_ver=_MIN_NUMEXPR_VERSION), UserWarning)
-
-except ImportError:  # pragma: no cover
-    pass
-
-__all__ = ['_NUMEXPR_INSTALLED']

pandas/core/computation/check.py

@@ -0,0 +1,22 @@
+import warnings
+from distutils.version import LooseVersion
+
+_NUMEXPR_INSTALLED = False
+_MIN_NUMEXPR_VERSION = "2.4.6"
+
+try:
+    import numexpr as ne
+    ver = ne.__version__
+    _NUMEXPR_INSTALLED = ver >= LooseVersion(_MIN_NUMEXPR_VERSION)
+
+    if not _NUMEXPR_INSTALLED:
+        warnings.warn(
+            "The installed version of numexpr {ver} is not supported "
+            "in pandas and will be not be used\nThe minimum supported "
+            "version is {min_ver}\n".format(
+                ver=ver, min_ver=_MIN_NUMEXPR_VERSION), UserWarning)
+
+except ImportError:  # pragma: no cover
+    pass
+
+__all__ = ['_NUMEXPR_INSTALLED']

pandas/core/computation/eval.py

@@ -5,8 +5,6 @@
 
 import tokenize
 from pandas.io.formats.printing import pprint_thing
-from pandas.core.computation import _NUMEXPR_INSTALLED
-from pandas.core.computation.expr import Expr, _parsers, tokenize_string
 from pandas.core.computation.scope import _ensure_scope
 from pandas.compat import string_types
 from pandas.core.computation.engines import _engines
@@ -32,6 +30,7 @@ def _check_engine(engine):
     string engine
 
     """
+    from pandas.core.computation.check import _NUMEXPR_INSTALLED
 
     if engine is None:
         if _NUMEXPR_INSTALLED:
@@ -69,6 +68,8 @@ def _check_parser(parser):
     KeyError
       * If an invalid parser is passed
     """
+    from pandas.core.computation.expr import _parsers
+
     if parser not in _parsers:
         raise KeyError('Invalid parser {parser!r} passed, valid parsers are'
                        ' {valid}'.format(parser=parser, valid=_parsers.keys()))
@@ -129,6 +130,8 @@ def _convert_expression(expr):
 
 
 def _check_for_locals(expr, stack_level, parser):
+    from pandas.core.computation.expr import tokenize_string
+
     at_top_of_stack = stack_level == 0
     not_pandas_parser = parser != 'pandas'
 
@@ -252,6 +255,7 @@ def eval(expr, parser='pandas', engine=None, truediv=True,
     pandas.DataFrame.query
     pandas.DataFrame.eval
     """
+    from pandas.core.computation.expr import Expr
 
     inplace = validate_bool_kwarg(inplace, "inplace")
 

pandas/core/computation/expressions.py

@@ -9,7 +9,7 @@
 import warnings
 import numpy as np
 from pandas.core.common import _values_from_object
-from pandas.core.computation import _NUMEXPR_INSTALLED
+from pandas.core.computation.check import _NUMEXPR_INSTALLED
 from pandas.core.config import get_option
 
 if _NUMEXPR_INSTALLED:

pandas/core/config_init.py

@@ -437,34 +437,36 @@ def use_inf_as_na_cb(key):
 writer_engine_doc = """
 : string
     The default Excel writer engine for '{ext}' files. Available options:
-    '{default}' (the default){others}.
-"""
-
-with cf.config_prefix('io.excel'):
-    # going forward, will be additional writers
-    for ext, options in [('xls', ['xlwt']), ('xlsm', ['openpyxl'])]:
-        default = options.pop(0)
-        if options:
-            options = " " + ", ".join(options)
-        else:
-            options = ""
-        doc = writer_engine_doc.format(ext=ext, default=default,
-                                       others=options)
-        cf.register_option(ext + '.writer', default, doc, validator=str)
-
-    def _register_xlsx(engine, other):
-        others = ", '{other}'".format(other=other)
-        doc = writer_engine_doc.format(ext='xlsx', default=engine,
-                                       others=others)
-        cf.register_option('xlsx.writer', engine, doc, validator=str)
-
-    try:
-        # better memory footprint
-        import xlsxwriter  # noqa
-        _register_xlsx('xlsxwriter', 'openpyxl')
-    except ImportError:
-        # fallback
-        _register_xlsx('openpyxl', 'xlsxwriter')
+    auto, {others}.
+"""
+
+_xls_options = ['xlwt']
+_xlsm_options = ['openpyxl']
+_xlsx_options = ['openpyxl', 'xlsxwriter']
+
+
+with cf.config_prefix("io.excel.xls"):
+    cf.register_option("writer", "auto",
+                       writer_engine_doc.format(
+                           ext='xls',
+                           others=', '.join(_xls_options)),
+                       validator=str)
+
+with cf.config_prefix("io.excel.xlsm"):
+    cf.register_option("writer", "auto",
+                       writer_engine_doc.format(
+                           ext='xlsm',
+                           others=', '.join(_xlsm_options)),
+                       validator=str)
+
+
+with cf.config_prefix("io.excel.xlsx"):
+    cf.register_option("writer", "auto",
+                       writer_engine_doc.format(
+                           ext='xlsx',
+                           others=', '.join(_xlsx_options)),
+                       validator=str)
+
 
 # Set up the io.parquet specific configuration.
 parquet_engine_doc = """

pandas/core/frame.py

@@ -76,9 +76,7 @@
                                    create_block_manager_from_blocks)
 from pandas.core.series import Series
 from pandas.core.categorical import Categorical
-import pandas.core.computation.expressions as expressions
 import pandas.core.algorithms as algorithms
-from pandas.core.computation.eval import eval as _eval
 from pandas.compat import (range, map, zip, lrange, lmap, lzip, StringIO, u,
                            OrderedDict, raise_with_traceback)
 from pandas import compat
@@ -2296,6 +2294,8 @@ def eval(self, expr, inplace=False, **kwargs):
         >>> df.eval('a + b')
         >>> df.eval('c = a + b')
         """
+        from pandas.core.computation.eval import eval as _eval
+
         inplace = validate_bool_kwarg(inplace, 'inplace')
         resolvers = kwargs.pop('resolvers', None)
         kwargs['level'] = kwargs.pop('level', 0) + 1
@@ -3840,6 +3840,7 @@ def _combine_const(self, other, func, raise_on_error=True, try_cast=True):
 
     def _compare_frame_evaluate(self, other, func, str_rep, try_cast=True):
 
+        import pandas.core.computation.expressions as expressions
         # unique
         if self.columns.is_unique:
 
@@ -3992,6 +3993,7 @@ def combine_first(self, other):
         -------
         combined : DataFrame
         """
+        import pandas.core.computation.expressions as expressions
 
         def combiner(x, y, needs_i8_conversion=False):
             x_values = x.values if hasattr(x, 'values') else x
@@ -4027,6 +4029,7 @@ def update(self, other, join='left', overwrite=True, filter_func=None,
             If True, will raise an error if the DataFrame and other both
             contain data in the same place.
         """
+        import pandas.core.computation.expressions as expressions
         # TODO: Support other joins
         if join != 'left':  # pragma: no cover
             raise NotImplementedError("Only left join is supported")

pandas/core/internals.py

@@ -66,7 +66,6 @@
 from pandas._libs.tslib import Timedelta
 from pandas._libs.lib import BlockPlacement
 
-import pandas.core.computation.expressions as expressions
 from pandas.util._decorators import cache_readonly
 from pandas.util._validators import validate_bool_kwarg
 from pandas import compat
@@ -1395,6 +1394,8 @@ def where(self, other, cond, align=True, raise_on_error=True,
         -------
         a new block(s), the result of the func
         """
+        import pandas.core.computation.expressions as expressions
+
         values = self.values
         orig_other = other
         if transpose:

pandas/core/ops.py

@@ -16,7 +16,6 @@
 
 from pandas import compat
 from pandas.util._decorators import Appender
-import pandas.core.computation.expressions as expressions
 
 from pandas.compat import bind_method
 import pandas.core.missing as missing
@@ -668,8 +667,9 @@ def _arith_method_SERIES(op, name, str_rep, fill_zeros=None, default_axis=None,
     Wrapper function for Series arithmetic operations, to avoid
     code duplication.
     """
-
     def na_op(x, y):
+        import pandas.core.computation.expressions as expressions
+
         try:
             result = expressions.evaluate(op, str_rep, x, y,
                                           raise_on_error=True, **eval_kwargs)
@@ -1193,6 +1193,8 @@ def to_series(right):
 def _arith_method_FRAME(op, name, str_rep=None, default_axis='columns',
                         fill_zeros=None, **eval_kwargs):
     def na_op(x, y):
+        import pandas.core.computation.expressions as expressions
+
         try:
             result = expressions.evaluate(op, str_rep, x, y,
                                           raise_on_error=True, **eval_kwargs)
@@ -1349,6 +1351,8 @@ def _arith_method_PANEL(op, name, str_rep=None, fill_zeros=None,
     # copied from Series na_op above, but without unnecessary branch for
     # non-scalar
     def na_op(x, y):
+        import pandas.core.computation.expressions as expressions
+
         try:
             result = expressions.evaluate(op, str_rep, x, y,
                                           raise_on_error=True, **eval_kwargs)
@@ -1378,6 +1382,8 @@ def f(self, other):
 
 def _comp_method_PANEL(op, name, str_rep=None, masker=False):
     def na_op(x, y):
+        import pandas.core.computation.expressions as expressions
+
         try:
             result = expressions.evaluate(op, str_rep, x, y,
                                           raise_on_error=True)

pandas/core/panel.py

@@ -15,7 +15,6 @@
     is_string_like, is_scalar)
 from pandas.core.dtypes.missing import notna
 
-import pandas.core.computation.expressions as expressions
 import pandas.core.common as com
 import pandas.core.ops as ops
 import pandas.core.missing as missing
@@ -1500,6 +1499,8 @@ def _add_aggregate_operations(cls, use_numexpr=True):
         def _panel_arith_method(op, name, str_rep=None, default_axis=None,
                                 fill_zeros=None, **eval_kwargs):
             def na_op(x, y):
+                import pandas.core.computation.expressions as expressions
+
                 try:
                     result = expressions.evaluate(op, str_rep, x, y,
                                                   raise_on_error=True,