Language Basics

Cython File Types

There are three file types in Cython:

  • Implementation files carry a .pyx suffix
  • Definition files carry a .pxd suffix
  • Include files which carry a .pxi suffix

Implementation File

What can it contain?

  • Basically anything Cythonic, but see below.

What can’t it contain?

  • There are some restrictions when it comes to extension types, if the extension type is already defined else where... more on this later

Definition File

What can it contain?

  • Any kind of C type declaration.
  • extern C function or variable declarations.
  • Declarations for module implementations.
  • The definition parts of extension types.
  • All declarations of functions, etc., for an external library

What can’t it contain?

  • Any non-extern C variable declaration.
  • Implementations of C or Python functions.
  • Python class definitions
  • Python executable statements.
  • Any declaration that is defined as public to make it accessible to other Cython modules.
  • This is not necessary, as it is automatic.
  • a public declaration is only needed to make it accessible to external C code.

What else?

  • Use the cimport statement, as you would Python’s import statement, to access these files from other definition or implementation files.
  • cimport does not need to be called in .pyx file for .pxd file that has the same name, as they are already in the same namespace.
  • For cimport to find the stated definition file, the path to the file must be appended to the -I option of the Cython compile command.
compilation order
  • When a .pyx file is to be compiled, Cython first checks to see if a corresponding .pxd file exits and processes it first.

Include File

What can it contain?

  • Any Cythonic code really, because the entire file is textually embedded at the location you prescribe.

How do I use it?

  • Include the .pxi file with an include statement like: include "spamstuff.pxi
  • The include statement can appear anywhere in your Cython file and at any indentation level
  • The code in the .pxi file needs to be rooted at the “zero” indentation level.
  • The included code can itself contain other include statements.

Declaring Data Types

As a dynamic language, Python encourages a programming style of considering classes and objects in terms of their methods and attributes, more than where they fit into the class hierarchy.

This can make Python a very relaxed and comfortable language for rapid development, but with a price - the ‘red tape’ of managing data types is dumped onto the interpreter. At run time, the interpreter does a lot of work searching namespaces, fetching attributes and parsing argument and keyword tuples. This run-time ‘late binding’ is a major cause of Python’s relative slowness compared to ‘early binding’ languages such as C++.

However with Cython it is possible to gain significant speed-ups through the use of ‘early binding’ programming techniques.


Typing is not a necessity

Providing static typing to parameters and variables is convenience to speed up your code, but it is not a necessity. Optimize where and when needed.

The cdef Statement

The cdef statement is used to make C level declarations for:

cdef int i, j, k
cdef float f, g[42], *h
cdef struct Grail:
    int age
    float volume

..note Structs can be declared as cdef packed struct, which has the same effect as the C directive #pragma pack(1).

cdef union Food:
    char *spam
    float *eggs
cdef enum CheeseType:
    cheddar, edam,

Declaring an enum as cpdef will create a PEP 435-style Python wrapper:

cpdef enum CheeseState:
    hard = 1
    soft = 2
    runny = 3
cdef int eggs(unsigned long l, float f):
Extension Types:
cdef class Spam:



Constants can be defined by using an anonymous enum:

cdef enum:
    tons_of_spam = 3

Grouping cdef Declarations

A series of declarations can grouped into a cdef block:

    struct Spam:
        int tons
    int i
    float f
    Spam *p
    void f(Spam *s):
    print s.tons, "Tons of spam"


ctypedef statement

The ctypedef statement is provided for naming types:

ctypedef unsigned long ULong
ctypedef int *IntPtr


  • Both C and Python function types can be declared to have parameters C data types.

  • Use normal C declaration syntax:

    def spam(int i, char *s):
        cdef int eggs(unsigned long l, float f):
  • As these parameters are passed into a Python declared function, they are magically converted to the specified C type value.

  • This holds true for only numeric and string types
  • If no type is specified for a parameter or a return value, it is assumed to be a Python object
  • The following takes two Python objects as parameters and returns a Python object:

    cdef spamobjs(x, y):


This is different then C language behavior, where it is an int by default.

  • Python object types have reference counting performed according to the standard Python C-API rules:
  • Borrowed references are taken as parameters
  • New references are returned


link or label here the one ref count caveat for NumPy.

  • The name object can be used to explicitly declare something as a Python Object.
  • For sake of code clarity, it recommended to always use object explicitly in your code.

  • This is also useful for cases where the name being declared would otherwise be taken for a type:

    cdef foo(object int):
  • As a return type:

    cdef object foo(object int):


Do a see also here ..??

Optional Arguments

  • Are supported for cdef and cpdef functions
  • There are differences though whether you declare them in a .pyx file or a .pxd file:
  • When in a .pyx file, the signature is the same as it is in Python itself:

    cdef class A:
        cdef foo(self):
            print "A"
    cdef class B(A)
        cdef foo(self, x=None)
            print "B", x
    cdef class C(B):
        cpdef foo(self, x=True, int k=3)
            print "C", x, k
  • When in a .pxd file, the signature is different like this example: cdef foo(x=*):

    cdef class A:
        cdef foo(self)
    cdef class B(A)
        cdef foo(self, x=*)
    cdef class C(B):
        cpdef foo(self, x=*, int k=*)
  • The number of arguments may increase when subclassing, but the arg types and order must be the same.
  • There may be a slight performance penalty when the optional arg is overridden with one that does not have default values.

Keyword-only Arguments

  • As in Python 3, def functions can have keyword-only arguments listed after a "*" parameter and before a "**" parameter if any:

    def f(a, b, *args, c, d = 42, e, **kwds):
  • Shown above, the c, d and e arguments can not be passed as positional arguments and must be passed as keyword arguments.
  • Furthermore, c and e are required keyword arguments since they do not have a default value.
  • If the parameter name after the "*" is omitted, the function will not accept any extra positional arguments:

    def g(a, b, *, c, d):
  • Shown above, the signature takes exactly two positional parameters and has two required keyword parameters

Automatic Type Conversion

  • For basic numeric and string types, in most situations, when a Python object is used in the context of a C value and vice versa.

  • The following table summarizes the conversion possibilities, assuming sizeof(int) == sizeof(long):

    C types From Python types To Python types
    [unsigned] char int, long int
    [unsigned] short
    int, long
    unsigned int int, long long
    unsigned long
    [unsigned] long long
    float, double, long double int, long, float float
    char * str/bytes str/bytes [1]
    struct   dict


Python String in a C Context

  • A Python string, passed to C context expecting a char*, is only valid as long as the Python string exists.

  • A reference to the Python string must be kept around for as long as the C string is needed.

  • If this can’t be guaranteed, then make a copy of the C string.

  • Cython may produce an error message: Obtaining char* from a temporary Python value and will not resume compiling in situations like this:

    cdef char *s
    s = pystring1 + pystring2
  • The reason is that concatenating to strings in Python produces a temporary variable.

  • The variable is decrefed, and the Python string deallocated as soon as the statement has finished,
  • Therefore the lvalue ``s`` is left dangling.
  • The solution is to assign the result of the concatenation to a Python variable, and then obtain the char* from that:

    cdef char *s
    p = pystring1 + pystring2
    s = p


It is up to you to be aware of this, and not to depend on Cython’s error message, as it is not guaranteed to be generated for every situation.

Type Casting

  • The syntax used in type casting are "<" and ">"


The syntax is different from C convention

cdef char *p, float *q
p = <char*>q
  • If one of the types is a python object for <type>x, Cython will try and do a coercion.


Cython will not stop a casting where there is no conversion, but it will emit a warning.

  • If the address is what is wanted, cast to a void* first.

Type Checking

  • A cast like <MyExtensionType>x will cast x to type MyExtensionType without type checking at all.
  • To have a cast type checked, use the syntax like: <MyExtensionType?>x.
  • In this case, Cython will throw an error if "x" is not a (subclass) of MyExtensionType
  • Automatic type checking for extension types can be obtained whenever isinstance() is used as the second parameter

Python Objects

Statements and Expressions

  • For the most part, control structures and expressions follow Python syntax.
  • When applied to Python objects, the semantics are the same unless otherwise noted.
  • Most Python operators can be applied to C values with the obvious semantics.
  • An expression with mixed Python and C values will have conversions performed automatically.
  • Python operations are automatically checked for errors, with the appropriate action taken.

Differences Between Cython and C

  • Most notable are C constructs which have no direct equivalent in Python.
  • An integer literal is treated as a C constant
  • It will be truncated to whatever size your C compiler thinks appropriate.

  • Cast to a Python object like this:

  • The "L", "LL" and the "U" suffixes have the same meaning as in C

  • There is no -> operator in Cython.. instead of p->x, use p.x.
  • There is no * operator in Cython.. instead of *p, use p[0].
  • & is permissible and has the same semantics as in C.
  • NULL is the null C pointer.
  • Do NOT use 0.
  • NULL is a reserved word in Cython
  • Syntax for Type casts are <type>value.

Scope Rules

  • All determination of scoping (local, module, built-in) in Cython is determined statically.
  • As with Python, a variable assignment which is not declared explicitly is implicitly declared to be a Python variable residing in the scope where it was assigned.


  • Module-level scope behaves the same way as a Python local scope if you refer to the variable before assigning to it.
  • Tricks, like the following will NOT work in Cython:

        x = True
    except NameError:
        True = 1
  • The above example will not work because True will always be looked up in the module-level scope. Do the following instead:

    import __builtin__
        True = __builtin__.True
    except AttributeError:
        True = 1

Built-in Constants

Predefined Python built-in constants:

  • None
  • True
  • False

Operator Precedence

  • Cython uses Python precedence order, not C


The “for ... in iterable” loop works as in Python, but is even more versatile in Cython as it can additionally be used on C types.

  • range() is C optimized when the index value has been declared by cdef, for example:

    cdef size_t i
    for i in range(n):
  • Iteration over C arrays and sliced pointers is supported and automatically infers the type of the loop variable, e.g.:

    cdef double* data = ...
    for x in data[:10]:
  • Iterating over many builtin types such as lists and tuples is optimized.

  • There is also a more verbose C-style for-from syntax which, however, is deprecated in favour of the normal Python “for ... in range()” loop. You might still find it in legacy code that was written for Pyrex, though.

  • The target expression must be a plain variable name.

  • The name between the lower and upper bounds must be the same as the target name.

    for i from 0 <= i < n:


  • Or when using a step size:

    for i from 0 <= i < n by s:
  • To reverse the direction, reverse the conditional operation:

    for i from n > i >= 0:
  • The break and continue statements are permissible.
  • Can contain an else clause.

Functions and Methods

  • There are three types of function declarations in Cython as the sub-sections show below.
  • Only “Python” functions can be called outside a Cython module from Python interpreted code.

Callable from Python (def)

  • Are declared with the def statement
  • Are called with Python objects
  • Return Python objects
  • See Parameters for special consideration

Callable from C (cdef)

  • Are declared with the cdef statement.
  • Are called with either Python objects or C values.
  • Can return either Python objects or C values.

Callable from both Python and C (cpdef)

  • Are declared with the cpdef statement.
  • Can be called from anywhere, because it uses a little Cython magic.
  • Uses the faster C calling conventions when being called from other Cython code.


cpdef methods can override cdef methods:

cdef class A:
    cdef foo(self):
        print "A"
cdef class B(A)
    cdef foo(self, x=None)
        print "B", x
cdef class C(B):
    cpdef foo(self, x=True, int k=3)
        print "C", x, k

When subclassing an extension type with a Python class, def methods can override cpdef methods but not cdef methods:

cdef class A:
    cdef foo(self):
cdef class B(A):
    cpdef foo(self):
class C(B):         # NOTE: not cdef class
    def foo(self):

If C above would be an extension type (cdef class), this would not work correctly. The Cython compiler will give a warning in that case.

Function Pointers

  • Functions declared in a struct are automatically converted to function pointers.
  • see using exceptions with function pointers

Python Built-ins

Cython compiles calls to most built-in functions into direct calls to the corresponding Python/C API routines, making them particularly fast.

Only direct function calls using these names are optimised. If you do something else with one of these names that assumes it’s a Python object, such as assign it to a Python variable, and later call it, the call will be made as a Python function call.

Function and arguments Return type Python/C API Equivalent
abs(obj) object, double, ... PyNumber_Absolute, fabs, fabsf, ...
callable(obj) bint PyObject_Callable
delattr(obj, name) None PyObject_DelAttr
exec(code, [glob, [loc]]) object
dir(obj) list PyObject_Dir
divmod(a, b) tuple PyNumber_Divmod
getattr(obj, name, [default]) (Note 1) object PyObject_GetAttr
hasattr(obj, name) bint PyObject_HasAttr
hash(obj) int / long PyObject_Hash
intern(obj) object Py*_InternFromString
isinstance(obj, type) bint PyObject_IsInstance
issubclass(obj, type) bint PyObject_IsSubclass
iter(obj, [sentinel]) object PyObject_GetIter
len(obj) Py_ssize_t PyObject_Length
pow(x, y, [z]) object PyNumber_Power
reload(obj) object PyImport_ReloadModule
repr(obj) object PyObject_Repr
setattr(obj, name) void PyObject_SetAttr

Note 1: Pyrex originally provided a function getattr3(obj, name, default)() corresponding to the three-argument form of the Python builtin getattr(). Cython still supports this function, but the usage is deprecated in favour of the normal builtin, which Cython can optimise in both forms.

Error and Exception Handling

  • A plain cdef declared function, that does not return a Python object...
  • Has no way of reporting a Python exception to it’s caller.
  • Will only print a warning message and the exception is ignored.
  • In order to propagate exceptions like this to it’s caller, you need to declare an exception value for it.
  • There are three forms of declaring an exception for a C compiled program.
  • First:

    cdef int spam() except -1:
  • In the example above, if an error occurs inside spam, it will immediately return with the value of -1, causing an exception to be propagated to it’s caller.
  • Functions declared with an exception value, should explicitly prevent a return of that value.
  • Second:

    cdef int spam() except? -1:
  • Used when a -1 may possibly be returned and is not to be considered an error.
  • The "?" tells Cython that -1 only indicates a possible error.
  • Now, each time -1 is returned, Cython generates a call to PyErr_Occurred to verify it is an actual error.
  • Third:

    cdef int spam() except *
  • A call to PyErr_Occurred happens every time the function gets called.


    Returning void

    A need to propagate errors when returning void must use this version.

  • Exception values can only be declared for functions returning an..
  • integer
  • enum
  • float
  • pointer type
  • Must be a constant expression



Function pointers

  • Require the same exception value specification as it’s user has declared.

  • Use cases here are when used as parameters and when assigned to a variable:

    int (*grail)(int, char *) except -1


Python Objects

  • Declared exception values are not need.
  • Remember that Cython assumes that a function without a declared return value, returns a Python object.
  • Exceptions on such functions are implicitly propagated by returning NULL



  • For exceptions from C++ compiled programs, see Wrapping C++ Classes

Checking return values for non-Cython functions..

  • Do not try to raise exceptions by returning the specified value.. Example:

    cdef extern FILE *fopen(char *filename, char *mode) except NULL # WRONG!
  • The except clause does not work that way.
  • It’s only purpose is to propagate Python exceptions that have already been raised by either...
  • A Cython function
  • A C function that calls Python/C API routines.
  • To propagate an exception for these circumstances you need to raise it yourself:

    cdef FILE *p
    p = fopen("spam.txt", "r")
    if p == NULL:
        raise SpamError("Couldn't open the spam file")

Conditional Compilation

  • The expressions in the following sub-sections must be valid compile-time expressions.
  • They can evaluate to any Python value.
  • The truth of the result is determined in the usual Python way.

Compile-Time Definitions

  • Defined using the DEF statement:

    DEF FavouriteFood = "spam"
    DEF ArraySize = 42
    DEF OtherArraySize = 2 * ArraySize + 17
  • The right hand side must be a valid compile-time expression made up of either:

  • Literal values
  • Names defined by other DEF statements
  • They can be combined using any of the Python expression syntax
  • Cython provides the following predefined names
  • Corresponding to the values returned by os.uname()
  • A name defined by DEF can appear anywhere an identifier can appear.
  • Cython replaces the name with the literal value before compilation.
  • The compile-time expression, in this case, must evaluate to a Python value of int, long, float, or str:

    cdef int a1[ArraySize]
    cdef int a2[OtherArraySize]
    print "I like", FavouriteFood

Conditional Statements

  • Similar semantics of the C pre-processor
  • The following statements can be used to conditionally include or exclude sections of code to compile.
  • IF
  • ELIF
  • ELSE
IF UNAME_SYSNAME == "Windows":
    include "icky_definitions.pxi"
    include "nice_definitions.pxi"
    include "penguin_definitions.pxi"
    include "other_definitions.pxi"
  • ELIF and ELSE are optional.
  • IF can appear anywhere that a normal statement or declaration can appear
  • It can contain any statements or declarations that would be valid in that context.
  • This includes other IF and DEF statements
[1]The conversion is to/from str for Python 2.x, and bytes for Python 3.x.