PageRenderTime 52ms CodeModel.GetById 7ms app.highlight 36ms RepoModel.GetById 1ms app.codeStats 1ms

ReStructuredText | 1017 lines | 769 code | 248 blank | 0 comment | 0 complexity | 267e4694c987b73b22e85cb1eb86ecc6 MD5 | raw file
   2.. _simple:
   5Simple statements
   8.. index:: pair: simple; statement
  10Simple statements are comprised within a single logical line. Several simple
  11statements may occur on a single line separated by semicolons.  The syntax for
  12simple statements is:
  14.. productionlist::
  15   simple_stmt: `expression_stmt`
  16              : | `assert_stmt`
  17              : | `assignment_stmt`
  18              : | `augmented_assignment_stmt`
  19              : | `pass_stmt`
  20              : | `del_stmt`
  21              : | `print_stmt`
  22              : | `return_stmt`
  23              : | `yield_stmt`
  24              : | `raise_stmt`
  25              : | `break_stmt`
  26              : | `continue_stmt`
  27              : | `import_stmt`
  28              : | `global_stmt`
  29              : | `exec_stmt`
  32.. _exprstmts:
  34Expression statements
  37.. index::
  38   pair: expression; statement
  39   pair: expression; list
  41Expression statements are used (mostly interactively) to compute and write a
  42value, or (usually) to call a procedure (a function that returns no meaningful
  43result; in Python, procedures return the value ``None``).  Other uses of
  44expression statements are allowed and occasionally useful.  The syntax for an
  45expression statement is:
  47.. productionlist::
  48   expression_stmt: `expression_list`
  50An expression statement evaluates the expression list (which may be a single
  53.. index::
  54   builtin: repr
  55   object: None
  56   pair: string; conversion
  57   single: output
  58   pair: standard; output
  59   pair: writing; values
  60   pair: procedure; call
  62In interactive mode, if the value is not ``None``, it is converted to a string
  63using the built-in :func:`repr` function and the resulting string is written to
  64standard output (see section :ref:`print`) on a line by itself.  (Expression
  65statements yielding ``None`` are not written, so that procedure calls do not
  66cause any output.)
  69.. _assignment:
  71Assignment statements
  74.. index::
  75   pair: assignment; statement
  76   pair: binding; name
  77   pair: rebinding; name
  78   object: mutable
  79   pair: attribute; assignment
  81Assignment statements are used to (re)bind names to values and to modify
  82attributes or items of mutable objects:
  84.. productionlist::
  85   assignment_stmt: (`target_list` "=")+ (`expression_list` | `yield_expression`)
  86   target_list: `target` ("," `target`)* [","]
  87   target: `identifier`
  88         : | "(" `target_list` ")"
  89         : | "[" `target_list` "]"
  90         : | `attributeref`
  91         : | `subscription`
  92         : | `slicing`
  94(See section :ref:`primaries` for the syntax definitions for the last three
  97.. index:: pair: expression; list
  99An assignment statement evaluates the expression list (remember that this can be
 100a single expression or a comma-separated list, the latter yielding a tuple) and
 101assigns the single resulting object to each of the target lists, from left to
 104.. index::
 105   single: target
 106   pair: target; list
 108Assignment is defined recursively depending on the form of the target (list).
 109When a target is part of a mutable object (an attribute reference, subscription
 110or slicing), the mutable object must ultimately perform the assignment and
 111decide about its validity, and may raise an exception if the assignment is
 112unacceptable.  The rules observed by various types and the exceptions raised are
 113given with the definition of the object types (see section :ref:`types`).
 115.. index:: triple: target; list; assignment
 117Assignment of an object to a target list is recursively defined as follows.
 119* If the target list is a single target: The object is assigned to that target.
 121* If the target list is a comma-separated list of targets: The object must be an
 122  iterable with the same number of items as there are targets in the target list,
 123  and the items are assigned, from left to right, to the corresponding targets.
 124  (This rule is relaxed as of Python 1.5; in earlier versions, the object had to
 125  be a tuple.  Since strings are sequences, an assignment like ``a, b = "xy"`` is
 126  now legal as long as the string has the right length.)
 128Assignment of an object to a single target is recursively defined as follows.
 130* If the target is an identifier (name):
 132    .. index:: statement: global
 134  * If the name does not occur in a :keyword:`global` statement in the current
 135    code block: the name is bound to the object in the current local namespace.
 137  * Otherwise: the name is bound to the object in the current global namespace.
 139  .. index:: single: destructor
 141  The name is rebound if it was already bound.  This may cause the reference count
 142  for the object previously bound to the name to reach zero, causing the object to
 143  be deallocated and its destructor (if it has one) to be called.
 145* If the target is a target list enclosed in parentheses or in square brackets:
 146  The object must be an iterable with the same number of items as there are
 147  targets in the target list, and its items are assigned, from left to right,
 148  to the corresponding targets.
 150  .. index:: pair: attribute; assignment
 152* If the target is an attribute reference: The primary expression in the
 153  reference is evaluated.  It should yield an object with assignable attributes;
 154  if this is not the case, :exc:`TypeError` is raised.  That object is then asked
 155  to assign the assigned object to the given attribute; if it cannot perform the
 156  assignment, it raises an exception (usually but not necessarily
 157  :exc:`AttributeError`).
 159  .. index::
 160     pair: subscription; assignment
 161     object: mutable
 163* If the target is a subscription: The primary expression in the reference is
 164  evaluated.  It should yield either a mutable sequence object (such as a list) or
 165  a mapping object (such as a dictionary). Next, the subscript expression is
 166  evaluated.
 168  .. index::
 169     object: sequence
 170     object: list
 172  If the primary is a mutable sequence object (such as a list), the subscript must
 173  yield a plain integer.  If it is negative, the sequence's length is added to it.
 174  The resulting value must be a nonnegative integer less than the sequence's
 175  length, and the sequence is asked to assign the assigned object to its item with
 176  that index.  If the index is out of range, :exc:`IndexError` is raised
 177  (assignment to a subscripted sequence cannot add new items to a list).
 179  .. index::
 180     object: mapping
 181     object: dictionary
 183  If the primary is a mapping object (such as a dictionary), the subscript must
 184  have a type compatible with the mapping's key type, and the mapping is then
 185  asked to create a key/datum pair which maps the subscript to the assigned
 186  object.  This can either replace an existing key/value pair with the same key
 187  value, or insert a new key/value pair (if no key with the same value existed).
 189  .. index:: pair: slicing; assignment
 191* If the target is a slicing: The primary expression in the reference is
 192  evaluated.  It should yield a mutable sequence object (such as a list).  The
 193  assigned object should be a sequence object of the same type.  Next, the lower
 194  and upper bound expressions are evaluated, insofar they are present; defaults
 195  are zero and the sequence's length.  The bounds should evaluate to (small)
 196  integers.  If either bound is negative, the sequence's length is added to it.
 197  The resulting bounds are clipped to lie between zero and the sequence's length,
 198  inclusive.  Finally, the sequence object is asked to replace the slice with the
 199  items of the assigned sequence.  The length of the slice may be different from
 200  the length of the assigned sequence, thus changing the length of the target
 201  sequence, if the object allows it.
 203(In the current implementation, the syntax for targets is taken to be the same
 204as for expressions, and invalid syntax is rejected during the code generation
 205phase, causing less detailed error messages.)
 207WARNING: Although the definition of assignment implies that overlaps between the
 208left-hand side and the right-hand side are 'safe' (for example ``a, b = b, a``
 209swaps two variables), overlaps *within* the collection of assigned-to variables
 210are not safe!  For instance, the following program prints ``[0, 2]``::
 212   x = [0, 1]
 213   i = 0
 214   i, x[i] = 1, 2
 215   print x
 218.. _augassign:
 220Augmented assignment statements
 223.. index::
 224   pair: augmented; assignment
 225   single: statement; assignment, augmented
 227Augmented assignment is the combination, in a single statement, of a binary
 228operation and an assignment statement:
 230.. productionlist::
 231   augmented_assignment_stmt: `augtarget` `augop` (`expression_list` | `yield_expression`)
 232   augtarget: `identifier` | `attributeref` | `subscription` | `slicing`
 233   augop: "+=" | "-=" | "*=" | "/=" | "//=" | "%=" | "**="
 234        : | ">>=" | "<<=" | "&=" | "^=" | "|="
 236(See section :ref:`primaries` for the syntax definitions for the last three
 239An augmented assignment evaluates the target (which, unlike normal assignment
 240statements, cannot be an unpacking) and the expression list, performs the binary
 241operation specific to the type of assignment on the two operands, and assigns
 242the result to the original target.  The target is only evaluated once.
 244An augmented assignment expression like ``x += 1`` can be rewritten as ``x = x +
 2451`` to achieve a similar, but not exactly equal effect. In the augmented
 246version, ``x`` is only evaluated once. Also, when possible, the actual operation
 247is performed *in-place*, meaning that rather than creating a new object and
 248assigning that to the target, the old object is modified instead.
 250With the exception of assigning to tuples and multiple targets in a single
 251statement, the assignment done by augmented assignment statements is handled the
 252same way as normal assignments. Similarly, with the exception of the possible
 253*in-place* behavior, the binary operation performed by augmented assignment is
 254the same as the normal binary operations.
 256For targets which are attribute references, the initial value is retrieved with
 257a :meth:`getattr` and the result is assigned with a :meth:`setattr`.  Notice
 258that the two methods do not necessarily refer to the same variable.  When
 259:meth:`getattr` refers to a class variable, :meth:`setattr` still writes to an
 260instance variable. For example::
 262   class A:
 263       x = 3    # class variable
 264   a = A()
 265   a.x += 1     # writes a.x as 4 leaving A.x as 3
 268.. _assert:
 270The :keyword:`assert` statement
 273.. index::
 274   statement: assert
 275   pair: debugging; assertions
 277Assert statements are a convenient way to insert debugging assertions into a
 280.. productionlist::
 281   assert_stmt: "assert" `expression` ["," `expression`]
 283The simple form, ``assert expression``, is equivalent to ::
 285   if __debug__:
 286      if not expression: raise AssertionError
 288The extended form, ``assert expression1, expression2``, is equivalent to ::
 290   if __debug__:
 291      if not expression1: raise AssertionError, expression2
 293.. index::
 294   single: __debug__
 295   exception: AssertionError
 297These equivalences assume that :const:`__debug__` and :exc:`AssertionError` refer to
 298the built-in variables with those names.  In the current implementation, the
 299built-in variable :const:`__debug__` is ``True`` under normal circumstances,
 300``False`` when optimization is requested (command line option -O).  The current
 301code generator emits no code for an assert statement when optimization is
 302requested at compile time.  Note that it is unnecessary to include the source
 303code for the expression that failed in the error message; it will be displayed
 304as part of the stack trace.
 306Assignments to :const:`__debug__` are illegal.  The value for the built-in variable
 307is determined when the interpreter starts.
 310.. _pass:
 312The :keyword:`pass` statement
 315.. index::
 316   statement: pass
 317   pair: null; operation
 319.. productionlist::
 320   pass_stmt: "pass"
 322:keyword:`pass` is a null operation --- when it is executed, nothing happens.
 323It is useful as a placeholder when a statement is required syntactically, but no
 324code needs to be executed, for example::
 326   def f(arg): pass    # a function that does nothing (yet)
 328   class C: pass       # a class with no methods (yet)
 331.. _del:
 333The :keyword:`del` statement
 336.. index::
 337   statement: del
 338   pair: deletion; target
 339   triple: deletion; target; list
 341.. productionlist::
 342   del_stmt: "del" `target_list`
 344Deletion is recursively defined very similar to the way assignment is defined.
 345Rather that spelling it out in full details, here are some hints.
 347Deletion of a target list recursively deletes each target, from left to right.
 349.. index::
 350   statement: global
 351   pair: unbinding; name
 353Deletion of a name removes the binding of that name  from the local or global
 354namespace, depending on whether the name occurs in a :keyword:`global` statement
 355in the same code block.  If the name is unbound, a :exc:`NameError` exception
 356will be raised.
 358.. index:: pair: free; variable
 360It is illegal to delete a name from the local namespace if it occurs as a free
 361variable in a nested block.
 363.. index:: pair: attribute; deletion
 365Deletion of attribute references, subscriptions and slicings is passed to the
 366primary object involved; deletion of a slicing is in general equivalent to
 367assignment of an empty slice of the right type (but even this is determined by
 368the sliced object).
 371.. _print:
 373The :keyword:`print` statement
 376.. index:: statement: print
 378.. productionlist::
 379   print_stmt: "print" ([`expression` ("," `expression`)* [","]]
 380             : | ">>" `expression` [("," `expression`)+ [","]])
 382:keyword:`print` evaluates each expression in turn and writes the resulting
 383object to standard output (see below).  If an object is not a string, it is
 384first converted to a string using the rules for string conversions.  The
 385(resulting or original) string is then written.  A space is written before each
 386object is (converted and) written, unless the output system believes it is
 387positioned at the beginning of a line.  This is the case (1) when no characters
 388have yet been written to standard output, (2) when the last character written to
 389standard output is a whitespace character except ``' '``, or (3) when the last
 390write operation on standard output was not a :keyword:`print` statement.
 391(In some cases it may be functional to write an empty string to standard output
 392for this reason.)
 394.. note::
 396   Objects which act like file objects but which are not the built-in file objects
 397   often do not properly emulate this aspect of the file object's behavior, so it
 398   is best not to rely on this.
 400.. index::
 401   single: output
 402   pair: writing; values
 403   pair: trailing; comma
 404   pair: newline; suppression
 406A ``'\n'`` character is written at the end, unless the :keyword:`print`
 407statement ends with a comma.  This is the only action if the statement contains
 408just the keyword :keyword:`print`.
 410.. index::
 411   pair: standard; output
 412   module: sys
 413   single: stdout (in module sys)
 414   exception: RuntimeError
 416Standard output is defined as the file object named ``stdout`` in the built-in
 417module :mod:`sys`.  If no such object exists, or if it does not have a
 418:meth:`write` method, a :exc:`RuntimeError` exception is raised.
 420.. index:: single: extended print statement
 422:keyword:`print` also has an extended form, defined by the second portion of the
 423syntax described above. This form is sometimes referred to as ":keyword:`print`
 424chevron." In this form, the first expression after the ``>>`` must evaluate to a
 425"file-like" object, specifically an object that has a :meth:`write` method as
 426described above.  With this extended form, the subsequent expressions are
 427printed to this file object.  If the first expression evaluates to ``None``,
 428then ``sys.stdout`` is used as the file for output.
 431.. _return:
 433The :keyword:`return` statement
 436.. index::
 437   statement: return
 438   pair: function; definition
 439   pair: class; definition
 441.. productionlist::
 442   return_stmt: "return" [`expression_list`]
 444:keyword:`return` may only occur syntactically nested in a function definition,
 445not within a nested class definition.
 447If an expression list is present, it is evaluated, else ``None`` is substituted.
 449:keyword:`return` leaves the current function call with the expression list (or
 450``None``) as return value.
 452.. index:: keyword: finally
 454When :keyword:`return` passes control out of a :keyword:`try` statement with a
 455:keyword:`finally` clause, that :keyword:`finally` clause is executed before
 456really leaving the function.
 458In a generator function, the :keyword:`return` statement is not allowed to
 459include an :token:`expression_list`.  In that context, a bare :keyword:`return`
 460indicates that the generator is done and will cause :exc:`StopIteration` to be
 464.. _yield:
 466The :keyword:`yield` statement
 469.. index::
 470   statement: yield
 471   single: generator; function
 472   single: generator; iterator
 473   single: function; generator
 474   exception: StopIteration
 476.. productionlist::
 477   yield_stmt: `yield_expression`
 479The :keyword:`yield` statement is only used when defining a generator function,
 480and is only used in the body of the generator function. Using a :keyword:`yield`
 481statement in a function definition is sufficient to cause that definition to
 482create a generator function instead of a normal function.
 484When a generator function is called, it returns an iterator known as a generator
 485iterator, or more commonly, a generator.  The body of the generator function is
 486executed by calling the generator's :meth:`next` method repeatedly until it
 487raises an exception.
 489When a :keyword:`yield` statement is executed, the state of the generator is
 490frozen and the value of :token:`expression_list` is returned to :meth:`next`'s
 491caller.  By "frozen" we mean that all local state is retained, including the
 492current bindings of local variables, the instruction pointer, and the internal
 493evaluation stack: enough information is saved so that the next time :meth:`next`
 494is invoked, the function can proceed exactly as if the :keyword:`yield`
 495statement were just another external call.
 497As of Python version 2.5, the :keyword:`yield` statement is now allowed in the
 498:keyword:`try` clause of a :keyword:`try` ...  :keyword:`finally` construct.  If
 499the generator is not resumed before it is finalized (by reaching a zero
 500reference count or by being garbage collected), the generator-iterator's
 501:meth:`close` method will be called, allowing any pending :keyword:`finally`
 502clauses to execute.
 504.. note::
 506   In Python 2.2, the :keyword:`yield` statement was only allowed when the
 507   ``generators`` feature has been enabled.  This ``__future__``
 508   import statement was used to enable the feature::
 510      from __future__ import generators
 513.. seealso::
 515   :pep:`0255` - Simple Generators
 516      The proposal for adding generators and the :keyword:`yield` statement to Python.
 518   :pep:`0342` - Coroutines via Enhanced Generators
 519      The proposal that, among other generator enhancements, proposed allowing
 520      :keyword:`yield` to appear inside a :keyword:`try` ... :keyword:`finally` block.
 523.. _raise:
 525The :keyword:`raise` statement
 528.. index::
 529   statement: raise
 530   single: exception
 531   pair: raising; exception
 533.. productionlist::
 534   raise_stmt: "raise" [`expression` ["," `expression` ["," `expression`]]]
 536If no expressions are present, :keyword:`raise` re-raises the last exception
 537that was active in the current scope.  If no exception is active in the current
 538scope, a :exc:`TypeError` exception is raised indicating that this is an error
 539(if running under IDLE, a :exc:`Queue.Empty` exception is raised instead).
 541Otherwise, :keyword:`raise` evaluates the expressions to get three objects,
 542using ``None`` as the value of omitted expressions.  The first two objects are
 543used to determine the *type* and *value* of the exception.
 545If the first object is an instance, the type of the exception is the class of
 546the instance, the instance itself is the value, and the second object must be
 549If the first object is a class, it becomes the type of the exception. The second
 550object is used to determine the exception value: If it is an instance of the
 551class, the instance becomes the exception value. If the second object is a
 552tuple, it is used as the argument list for the class constructor; if it is
 553``None``, an empty argument list is used, and any other object is treated as a
 554single argument to the constructor.  The instance so created by calling the
 555constructor is used as the exception value.
 557.. index:: object: traceback
 559If a third object is present and not ``None``, it must be a traceback object
 560(see section :ref:`types`), and it is substituted instead of the current
 561location as the place where the exception occurred.  If the third object is
 562present and not a traceback object or ``None``, a :exc:`TypeError` exception is
 563raised.  The three-expression form of :keyword:`raise` is useful to re-raise an
 564exception transparently in an except clause, but :keyword:`raise` with no
 565expressions should be preferred if the exception to be re-raised was the most
 566recently active exception in the current scope.
 568Additional information on exceptions can be found in section :ref:`exceptions`,
 569and information about handling exceptions is in section :ref:`try`.
 572.. _break:
 574The :keyword:`break` statement
 577.. index::
 578   statement: break
 579   statement: for
 580   statement: while
 581   pair: loop; statement
 583.. productionlist::
 584   break_stmt: "break"
 586:keyword:`break` may only occur syntactically nested in a :keyword:`for` or
 587:keyword:`while` loop, but not nested in a function or class definition within
 588that loop.
 590.. index:: keyword: else
 592It terminates the nearest enclosing loop, skipping the optional :keyword:`else`
 593clause if the loop has one.
 595.. index:: pair: loop control; target
 597If a :keyword:`for` loop is terminated by :keyword:`break`, the loop control
 598target keeps its current value.
 600.. index:: keyword: finally
 602When :keyword:`break` passes control out of a :keyword:`try` statement with a
 603:keyword:`finally` clause, that :keyword:`finally` clause is executed before
 604really leaving the loop.
 607.. _continue:
 609The :keyword:`continue` statement
 612.. index::
 613   statement: continue
 614   statement: for
 615   statement: while
 616   pair: loop; statement
 617   keyword: finally
 619.. productionlist::
 620   continue_stmt: "continue"
 622:keyword:`continue` may only occur syntactically nested in a :keyword:`for` or
 623:keyword:`while` loop, but not nested in a function or class definition or
 624:keyword:`finally` clause within that loop.  It continues with the next
 625cycle of the nearest enclosing loop.
 627When :keyword:`continue` passes control out of a :keyword:`try` statement with a
 628:keyword:`finally` clause, that :keyword:`finally` clause is executed before
 629really starting the next loop cycle.
 632.. _import:
 633.. _from:
 635The :keyword:`import` statement
 638.. index::
 639   statement: import
 640   single: module; importing
 641   pair: name; binding
 642   keyword: from
 644.. productionlist::
 645   import_stmt: "import" `module` ["as" `name`] ( "," `module` ["as" `name`] )*
 646              : | "from" `relative_module` "import" `identifier` ["as" `name`]
 647              : ( "," `identifier` ["as" `name`] )*
 648              : | "from" `relative_module` "import" "(" `identifier` ["as" `name`]
 649              : ( "," `identifier` ["as" `name`] )* [","] ")"
 650              : | "from" `module` "import" "*"
 651   module: (`identifier` ".")* `identifier`
 652   relative_module: "."* `module` | "."+
 653   name: `identifier`
 655Import statements are executed in two steps: (1) find a module, and initialize
 656it if necessary; (2) define a name or names in the local namespace (of the scope
 657where the :keyword:`import` statement occurs). The statement comes in two
 658forms differing on whether it uses the :keyword:`from` keyword. The first form
 659(without :keyword:`from`) repeats these steps for each identifier in the list.
 660The form with :keyword:`from` performs step (1) once, and then performs step
 661(2) repeatedly.
 663.. index::
 664    single: package
 666To understand how step (1) occurs, one must first understand how Python handles
 667hierarchical naming of modules. To help organize modules and provide a
 668hierarchy in naming, Python has a concept of packages. A package can contain
 669other packages and modules while modules cannot contain other modules or
 670packages. From a file system perspective, packages are directories and modules
 671are files. The original `specification for packages
 672<>`_ is still available to read,
 673although minor details have changed since the writing of that document.
 675.. index::
 676    single: sys.modules
 678Once the name of the module is known (unless otherwise specified, the term
 679"module" will refer to both packages and modules), searching
 680for the module or package can begin. The first place checked is
 681:data:`sys.modules`, the cache of all modules that have been imported
 682previously. If the module is found there then it is used in step (2) of import.
 684.. index::
 685    single: sys.meta_path
 686    single: finder
 687    pair: finder; find_module
 688    single: __path__
 690If the module is not found in the cache, then :data:`sys.meta_path` is searched
 691(the specification for :data:`sys.meta_path` can be found in :pep:`302`).
 692The object is a list of :term:`finder` objects which are queried in order as to
 693whether they know how to load the module by calling their :meth:`find_module`
 694method with the name of the module. If the module happens to be contained
 695within a package (as denoted by the existence of a dot in the name), then a
 696second argument to :meth:`find_module` is given as the value of the
 697:attr:`__path__` attribute from the parent package (everything up to the last
 698dot in the name of the module being imported). If a finder can find the module
 699it returns a :term:`loader` (discussed later) or returns :keyword:`None`.
 701.. index::
 702    single: sys.path_hooks
 703    single: sys.path_importer_cache
 704    single: sys.path
 706If none of the finders on :data:`sys.meta_path` are able to find the module
 707then some implicitly defined finders are queried. Implementations of Python
 708vary in what implicit meta path finders are defined. The one they all do
 709define, though, is one that handles :data:`sys.path_hooks`,
 710:data:`sys.path_importer_cache`, and :data:`sys.path`.
 712The implicit finder searches for the requested module in the "paths" specified
 713in one of two places ("paths" do not have to be file system paths). If the
 714module being imported is supposed to be contained within a package then the
 715second argument passed to :meth:`find_module`, :attr:`__path__` on the parent
 716package, is used as the source of paths. If the module is not contained in a
 717package then :data:`sys.path` is used as the source of paths.
 719Once the source of paths is chosen it is iterated over to find a finder that
 720can handle that path. The dict at :data:`sys.path_importer_cache` caches
 721finders for paths and is checked for a finder. If the path does not have a
 722finder cached then :data:`sys.path_hooks` is searched by calling each object in
 723the list with a single argument of the path, returning a finder or raises
 724:exc:`ImportError`. If a finder is returned then it is cached in
 725:data:`sys.path_importer_cache` and then used for that path entry. If no finder
 726can be found but the path exists then a value of :keyword:`None` is
 727stored in :data:`sys.path_importer_cache` to signify that an implicit,
 728file-based finder that handles modules stored as individual files should be
 729used for that path. If the path does not exist then a finder which always
 730returns :keyword:`None` is placed in the cache for the path.
 732.. index::
 733    single: loader
 734    pair: loader; load_module
 735    exception: ImportError
 737If no finder can find the module then :exc:`ImportError` is raised. Otherwise
 738some finder returned a loader whose :meth:`load_module` method is called with
 739the name of the module to load (see :pep:`302` for the original definition of
 740loaders). A loader has several responsibilities to perform on a module it
 741loads. First, if the module already exists in :data:`sys.modules` (a
 742possibility if the loader is called outside of the import machinery) then it
 743is to use that module for initialization and not a new module. But if the
 744module does not exist in :data:`sys.modules` then it is to be added to that
 745dict before initialization begins. If an error occurs during loading of the
 746module and it was added to :data:`sys.modules` it is to be removed from the
 747dict. If an error occurs but the module was already in :data:`sys.modules` it
 748is left in the dict.
 750.. index::
 751    single: __name__
 752    single: __file__
 753    single: __path__
 754    single: __package__
 755    single: __loader__
 757The loader must set several attributes on the module. :data:`__name__` is to be
 758set to the name of the module. :data:`__file__` is to be the "path" to the file
 759unless the module is built-in (and thus listed in
 760:data:`sys.builtin_module_names`) in which case the attribute is not set.
 761If what is being imported is a package then :data:`__path__` is to be set to a
 762list of paths to be searched when looking for modules and packages contained
 763within the package being imported. :data:`__package__` is optional but should
 764be set to the name of package that contains the module or package (the empty
 765string is used for module not contained in a package). :data:`__loader__` is
 766also optional but should be set to the loader object that is loading the
 769.. index::
 770    exception: ImportError
 772If an error occurs during loading then the loader raises :exc:`ImportError` if
 773some other exception is not already being propagated. Otherwise the loader
 774returns the module that was loaded and initialized.
 776When step (1) finishes without raising an exception, step (2) can begin.
 778The first form of :keyword:`import` statement binds the module name in the local
 779namespace to the module object, and then goes on to import the next identifier,
 780if any.  If the module name is followed by :keyword:`as`, the name following
 781:keyword:`as` is used as the local name for the module.
 783.. index::
 784   pair: name; binding
 785   exception: ImportError
 787The :keyword:`from` form does not bind the module name: it goes through the list
 788of identifiers, looks each one of them up in the module found in step (1), and
 789binds the name in the local namespace to the object thus found.  As with the
 790first form of :keyword:`import`, an alternate local name can be supplied by
 791specifying ":keyword:`as` localname".  If a name is not found,
 792:exc:`ImportError` is raised.  If the list of identifiers is replaced by a star
 793(``'*'``), all public names defined in the module are bound in the local
 794namespace of the :keyword:`import` statement..
 796.. index:: single: __all__ (optional module attribute)
 798The *public names* defined by a module are determined by checking the module's
 799namespace for a variable named ``__all__``; if defined, it must be a sequence of
 800strings which are names defined or imported by that module.  The names given in
 801``__all__`` are all considered public and are required to exist.  If ``__all__``
 802is not defined, the set of public names includes all names found in the module's
 803namespace which do not begin with an underscore character (``'_'``).
 804``__all__`` should contain the entire public API. It is intended to avoid
 805accidentally exporting items that are not part of the API (such as library
 806modules which were imported and used within the module).
 808The :keyword:`from` form with ``*`` may only occur in a module scope.  If the
 809wild card form of import --- ``import *`` --- is used in a function and the
 810function contains or is a nested block with free variables, the compiler will
 811raise a :exc:`SyntaxError`.
 813.. index::
 814    single: relative; import
 816When specifying what module to import you do not have to specify the absolute
 817name of the module. When a module or package is contained within another
 818package it is possible to make a relative import within the same top package
 819without having to mention the package name. By using leading dots in the
 820specified module or package after :keyword:`from` you can specify how high to
 821traverse up the current package hierarchy without specifying exact names. One
 822leading dot means the current package where the module making the import
 823exists. Two dots means up one package level. Three dots is up two levels, etc.
 824So if you execute ``from . import mod`` from a module in the ``pkg`` package
 825then you will end up importing ``pkg.mod``. If you execute ``from ..subpkg2
 826imprt mod`` from within ``pkg.subpkg1`` you will import ``pkg.subpkg2.mod``.
 827The specification for relative imports is contained within :pep:`328`.
 830.. index:: builtin: __import__
 832The built-in function :func:`__import__` is provided to support applications
 833that determine which modules need to be loaded dynamically; refer to
 834:ref:`built-in-funcs` for additional information.
 837.. _future:
 839Future statements
 842.. index:: pair: future; statement
 844A :dfn:`future statement` is a directive to the compiler that a particular
 845module should be compiled using syntax or semantics that will be available in a
 846specified future release of Python.  The future statement is intended to ease
 847migration to future versions of Python that introduce incompatible changes to
 848the language.  It allows use of the new features on a per-module basis before
 849the release in which the feature becomes standard.
 851.. productionlist:: *
 852   future_statement: "from" "__future__" "import" feature ["as" name]
 853                   : ("," feature ["as" name])*
 854                   : | "from" "__future__" "import" "(" feature ["as" name]
 855                   : ("," feature ["as" name])* [","] ")"
 856   feature: identifier
 857   name: identifier
 859A future statement must appear near the top of the module.  The only lines that
 860can appear before a future statement are:
 862* the module docstring (if any),
 863* comments,
 864* blank lines, and
 865* other future statements.
 867The features recognized by Python 2.6 are ``unicode_literals``,
 868``print_function``, ``absolute_import``, ``division``, ``generators``,
 869``nested_scopes`` and ``with_statement``.  ``generators``, ``with_statement``,
 870``nested_scopes`` are redundant in Python version 2.6 and above because they are
 871always enabled.
 873A future statement is recognized and treated specially at compile time: Changes
 874to the semantics of core constructs are often implemented by generating
 875different code.  It may even be the case that a new feature introduces new
 876incompatible syntax (such as a new reserved word), in which case the compiler
 877may need to parse the module differently.  Such decisions cannot be pushed off
 878until runtime.
 880For any given release, the compiler knows which feature names have been defined,
 881and raises a compile-time error if a future statement contains a feature not
 882known to it.
 884The direct runtime semantics are the same as for any import statement: there is
 885a standard module :mod:`__future__`, described later, and it will be imported in
 886the usual way at the time the future statement is executed.
 888The interesting runtime semantics depend on the specific feature enabled by the
 889future statement.
 891Note that there is nothing special about the statement::
 893   import __future__ [as name]
 895That is not a future statement; it's an ordinary import statement with no
 896special semantics or syntax restrictions.
 898Code compiled by an :keyword:`exec` statement or calls to the builtin functions
 899:func:`compile` and :func:`execfile` that occur in a module :mod:`M` containing
 900a future statement will, by default, use the new  syntax or semantics associated
 901with the future statement.  This can, starting with Python 2.2 be controlled by
 902optional arguments to :func:`compile` --- see the documentation of that function
 903for details.
 905A future statement typed at an interactive interpreter prompt will take effect
 906for the rest of the interpreter session.  If an interpreter is started with the
 907:option:`-i` option, is passed a script name to execute, and the script includes
 908a future statement, it will be in effect in the interactive session started
 909after the script is executed.
 911.. seealso::
 913   :pep:`236` - Back to the __future__
 914      The original proposal for the __future__ mechanism.
 917.. _global:
 919The :keyword:`global` statement
 922.. index::
 923   statement: global
 924   triple: global; name; binding
 926.. productionlist::
 927   global_stmt: "global" `identifier` ("," `identifier`)*
 929The :keyword:`global` statement is a declaration which holds for the entire
 930current code block.  It means that the listed identifiers are to be interpreted
 931as globals.  It would be impossible to assign to a global variable without
 932:keyword:`global`, although free variables may refer to globals without being
 933declared global.
 935Names listed in a :keyword:`global` statement must not be used in the same code
 936block textually preceding that :keyword:`global` statement.
 938Names listed in a :keyword:`global` statement must not be defined as formal
 939parameters or in a :keyword:`for` loop control target, :keyword:`class`
 940definition, function definition, or :keyword:`import` statement.
 942(The current implementation does not enforce the latter two restrictions, but
 943programs should not abuse this freedom, as future implementations may enforce
 944them or silently change the meaning of the program.)
 946.. index::
 947   statement: exec
 948   builtin: eval
 949   builtin: execfile
 950   builtin: compile
 952**Programmer's note:** the :keyword:`global` is a directive to the parser.  It
 953applies only to code parsed at the same time as the :keyword:`global` statement.
 954In particular, a :keyword:`global` statement contained in an :keyword:`exec`
 955statement does not affect the code block *containing* the :keyword:`exec`
 956statement, and code contained in an :keyword:`exec` statement is unaffected by
 957:keyword:`global` statements in the code containing the :keyword:`exec`
 958statement.  The same applies to the :func:`eval`, :func:`execfile` and
 959:func:`compile` functions.
 962.. _exec:
 964The :keyword:`exec` statement
 967.. index:: statement: exec
 969.. productionlist::
 970   exec_stmt: "exec" `or_expr` ["in" `expression` ["," `expression`]]
 972This statement supports dynamic execution of Python code.  The first expression
 973should evaluate to either a string, an open file object, or a code object.  If
 974it is a string, the string is parsed as a suite of Python statements which is
 975then executed (unless a syntax error occurs). [#]_  If it is an open file, the file
 976is parsed until EOF and executed.  If it is a code object, it is simply
 977executed.  In all cases, the code that's executed is expected to be valid as
 978file input (see section :ref:`file-input`).  Be aware that the
 979:keyword:`return` and :keyword:`yield` statements may not be used outside of
 980function definitions even within the context of code passed to the
 981:keyword:`exec` statement.
 983In all cases, if the optional parts are omitted, the code is executed in the
 984current scope.  If only the first expression after :keyword:`in` is specified,
 985it should be a dictionary, which will be used for both the global and the local
 986variables.  If two expressions are given, they are used for the global and local
 987variables, respectively. If provided, *locals* can be any mapping object.
 989.. versionchanged:: 2.4
 990   Formerly, *locals* was required to be a dictionary.
 992.. index::
 993   single: __builtins__
 994   module: __builtin__
 996As a side effect, an implementation may insert additional keys into the
 997dictionaries given besides those corresponding to variable names set by the
 998executed code.  For example, the current implementation may add a reference to
 999the dictionary of the built-in module :mod:`__builtin__` under the key
1000``__builtins__`` (!).
1002.. index::
1003   builtin: eval
1004   builtin: globals
1005   builtin: locals
1007**Programmer's hints:** dynamic evaluation of expressions is supported by the
1008built-in function :func:`eval`.  The built-in functions :func:`globals` and
1009:func:`locals` return the current global and local dictionary, respectively,
1010which may be useful to pass around for use by :keyword:`exec`.
1013.. rubric:: Footnotes
1015.. [#] Note that the parser only accepts the Unix-style end of line convention.
1016       If you are reading the code from a file, make sure to use universal
1017       newline mode to convert Windows or Mac-style newlines.