warnings — Warning control¶

Warning Categories¶

There are a number of built-in exceptions that represent warning categories. This categorization is useful to be able to filter out groups of warnings.

While these are technically built-in exceptions, they are documented here, because conceptually they belong to the warnings mechanism.

User code can define additional warning categories by subclassing one of the standard warning categories. A warning category must always be a subclass of the Warning class.

The following warnings category classes are currently defined:

The Warnings Filter¶

The warnings filter controls whether warnings are ignored, displayed, or turned into errors (raising an exception).

Conceptually, the warnings filter maintains an ordered list of filter specifications; any specific warning is matched against each filter specification in the list in turn until a match is found; the filter determines the disposition of the match. Each entry is a tuple of the form (action, message, category, module, lineno), where:

• action is one of the following strings:

  Value	Disposition
  "default"	print the first occurrence of matching warnings for each location (module + line number) where the warning is issued
  "error"	turn matching warnings into exceptions
  "ignore"	never print matching warnings
  "always"	always print matching warnings
  "all"	alias to “always”
  "module"	print the first occurrence of matching warnings for each module where the warning is issued (regardless of line number)
  "once"	print only the first occurrence of matching warnings, regardless of location

• message is a string containing a regular expression that the start of the warning message must match, case-insensitively. In -W and PYTHONWARNINGS, message is a literal string that the start of the warning message must contain (case-insensitively), ignoring any whitespace at the start or end of message.

• category is a class (a subclass of Warning) of which the warning category must be a subclass in order to match.

• module is a string containing a regular expression that the start of the fully qualified module name must match, case-sensitively. In -W and PYTHONWARNINGS, module is a literal string that the fully qualified module name must be equal to (case-sensitively), ignoring any whitespace at the start or end of module.

• lineno is an integer that the line number where the warning occurred must match, or 0 to match all line numbers.

Since the Warning class is derived from the built-in Exception class, to turn a warning into an error we simply raise category(message).

If a warning is reported and doesn’t match any registered filter then the “default” action is applied (hence its name).

action:message:category:module:line

default                      # Show all warnings (even those ignored by default)
ignore                       # Ignore all warnings
error                        # Convert all warnings to errors
error::ResourceWarning       # Treat ResourceWarning messages as errors
default::DeprecationWarning  # Show DeprecationWarning messages
ignore,default:::mymodule    # Only report warnings triggered by "mymodule"
error:::mymodule             # Convert warnings to errors in "mymodule"

default::DeprecationWarning:__main__
ignore::DeprecationWarning
ignore::PendingDeprecationWarning
ignore::ImportWarning
ignore::ResourceWarning

import sys

if not sys.warnoptions:
    import warnings
    warnings.simplefilter("ignore")

import sys

if not sys.warnoptions:
    import os, warnings
    warnings.simplefilter("default") # Change the filter in this process
    os.environ["PYTHONWARNINGS"] = "default" # Also affect subprocesses

import warnings
warnings.filterwarnings("default", category=DeprecationWarning,
                                   module=user_ns.get("__name__"))

Temporarily Suppressing Warnings¶

If you are using code that you know will raise a warning, such as a deprecated function, but do not want to see the warning (even when warnings have been explicitly configured via the command line), then it is possible to suppress the warning using the catch_warnings context manager:

import warnings

def fxn():
    warnings.warn("deprecated", DeprecationWarning)

with warnings.catch_warnings():
    warnings.simplefilter("ignore")
    fxn()

While within the context manager all warnings will simply be ignored. This allows you to use known-deprecated code without having to see the warning while not suppressing the warning for other code that might not be aware of its use of deprecated code.

Note

See Concurrent safety of Context Managers for details on the concurrency-safety of the catch_warnings context manager when used in programs using multiple threads or async functions.

Testing Warnings¶

To test warnings raised by code, use the catch_warnings context manager. With it you can temporarily mutate the warnings filter to facilitate your testing. For instance, do the following to capture all raised warnings to check:

import warnings

def fxn():
    warnings.warn("deprecated", DeprecationWarning)

with warnings.catch_warnings(record=True) as w:
    # Cause all warnings to always be triggered.
    warnings.simplefilter("always")
    # Trigger a warning.
    fxn()
    # Verify some things
    assert len(w) == 1
    assert issubclass(w[-1].category, DeprecationWarning)
    assert "deprecated" in str(w[-1].message)

One can also cause all warnings to be exceptions by using error instead of always. One thing to be aware of is that if a warning has already been raised because of a once/default rule, then no matter what filters are set the warning will not be seen again unless the warnings registry related to the warning has been cleared.

Once the context manager exits, the warnings filter is restored to its state when the context was entered. This prevents tests from changing the warnings filter in unexpected ways between tests and leading to indeterminate test results.

Note

See Concurrent safety of Context Managers for details on the concurrency-safety of the catch_warnings context manager when used in programs using multiple threads or async functions.

When testing multiple operations that raise the same kind of warning, it is important to test them in a manner that confirms each operation is raising a new warning (e.g. set warnings to be raised as exceptions and check the operations raise exceptions, check that the length of the warning list continues to increase after each operation, or else delete the previous entries from the warnings list before each new operation).

Updating Code For New Versions of Dependencies¶

Warning categories that are primarily of interest to Python developers (rather than end users of applications written in Python) are ignored by default.

Notably, this “ignored by default” list includes DeprecationWarning (for every module except __main__), which means developers should make sure to test their code with typically ignored warnings made visible in order to receive timely notifications of future breaking API changes (whether in the standard library or third party packages).

In the ideal case, the code will have a suitable test suite, and the test runner will take care of implicitly enabling all warnings when running tests (the test runner provided by the unittest module does this).

In less ideal cases, applications can be checked for use of deprecated interfaces by passing -Wd to the Python interpreter (this is shorthand for -W default) or setting PYTHONWARNINGS=default in the environment. This enables default handling for all warnings, including those that are ignored by default. To change what action is taken for encountered warnings you can change what argument is passed to -W (e.g. -W error). See the -W flag for more details on what is possible.

Available Functions¶

warnings.warn(message, category=None, stacklevel=1, source=None, *, skip_file_prefixes=())¶

Issue a warning, or maybe ignore it or raise an exception. The category argument, if given, must be a warning category class; it defaults to UserWarning. Alternatively, message can be a Warning instance, in which case category will be ignored and message.__class__ will be used. In this case, the message text will be str(message). This function raises an exception if the particular warning issued is changed into an error by the warnings filter. The stacklevel argument can be used by wrapper functions written in Python, like this:

def deprecated_api(message):
    warnings.warn(message, DeprecationWarning, stacklevel=2)

This makes the warning refer to deprecated_api’s caller, rather than to the source of deprecated_api itself (since the latter would defeat the purpose of the warning message).

The skip_file_prefixes keyword argument can be used to indicate which stack frames are ignored when counting stack levels. This can be useful when you want the warning to always appear at call sites outside of a package when a constant stacklevel does not fit all call paths or is otherwise challenging to maintain. If supplied, it must be a tuple of strings. When prefixes are supplied, stacklevel is implicitly overridden to be max(2, stacklevel). To cause a warning to be attributed to the caller from outside of the current package you might write:

# example/lower.py
_warn_skips = (os.path.dirname(__file__),)

def one_way(r_luxury_yacht=None, t_wobbler_mangrove=None):
    if r_luxury_yacht:
        warnings.warn("Please migrate to t_wobbler_mangrove=.",
                      skip_file_prefixes=_warn_skips)

# example/higher.py
from . import lower

def another_way(**kw):
    lower.one_way(**kw)

This makes the warning refer to both the example.lower.one_way() and example.higher.another_way() call sites only from calling code living outside of example package.

source, if supplied, is the destroyed object which emitted a ResourceWarning.

Changed in version 3.6: Added source parameter.

Changed in version 3.12: Added skip_file_prefixes.

warnings.warn_explicit(message, category, filename, lineno, module=None, registry=None, module_globals=None, source=None)¶

This is a low-level interface to the functionality of warn(), passing in explicitly the message, category, filename and line number, and optionally other arguments. message must be a string and category a subclass of Warning or message may be a Warning instance, in which case category will be ignored.

module, if supplied, should be the module name. If no module is passed, the filename with .py stripped is used.

registry, if supplied, should be the __warningregistry__ dictionary of the module. If no registry is passed, each warning is treated as the first occurrence, that is, filter actions "default", "module" and "once" are handled as "always".

module_globals, if supplied, should be the global namespace in use by the code for which the warning is issued. (This argument is used to support displaying source for modules found in zipfiles or other non-filesystem import sources).

source, if supplied, is the destroyed object which emitted a ResourceWarning.

Changed in version 3.6: Add the source parameter.

warnings.showwarning(message, category, filename, lineno, file=None, line=None)¶

Write a warning to a file. The default implementation calls formatwarning(message, category, filename, lineno, line) and writes the resulting string to file, which defaults to sys.stderr. You may replace this function with any callable by assigning to warnings.showwarning. line is a line of source code to be included in the warning message; if line is not supplied, showwarning() will try to read the line specified by filename and lineno.

warnings.formatwarning(message, category, filename, lineno, line=None)¶

Format a warning the standard way. This returns a string which may contain embedded newlines and ends in a newline. line is a line of source code to be included in the warning message; if line is not supplied, formatwarning() will try to read the line specified by filename and lineno.

warnings.filterwarnings(action, message='', category=Warning, module='', lineno=0, append=False)¶

Insert an entry into the list of warnings filter specifications. The entry is inserted at the front by default; if append is true, it is inserted at the end. This checks the types of the arguments, compiles the message and module regular expressions, and inserts them as a tuple in the list of warnings filters. Entries closer to the front of the list override entries later in the list, if both match a particular warning. Omitted arguments default to a value that matches everything.

warnings.simplefilter(action, category=Warning, lineno=0, append=False)¶

Insert a simple entry into the list of warnings filter specifications. The meaning of the function parameters is as for filterwarnings(), but regular expressions are not needed as the filter inserted always matches any message in any module as long as the category and line number match.

warnings.resetwarnings()¶

Reset the warnings filter. This discards the effect of all previous calls to filterwarnings(), including that of the -W command line options and calls to simplefilter().

@warnings.deprecated(message, /, *, category=DeprecationWarning, stacklevel=1)¶

Decorator to indicate that a class, function or overload is deprecated.

When this decorator is applied to an object, deprecation warnings may be emitted at runtime when the object is used. static type checkers will also generate a diagnostic on usage of the deprecated object.

Usage:

from warnings import deprecated
from typing import overload

@deprecated("Use B instead")
class A:
    pass

@deprecated("Use g instead")
def f():
    pass

@overload
@deprecated("int support is deprecated")
def g(x: int) -> int: ...
@overload
def g(x: str) -> int: ...

The warning specified by category will be emitted at runtime on use of deprecated objects. For functions, that happens on calls; for classes, on instantiation and on creation of subclasses. If the category is None, no warning is emitted at runtime. The stacklevel determines where the warning is emitted. If it is 1 (the default), the warning is emitted at the direct caller of the deprecated object; if it is higher, it is emitted further up the stack. Static type checker behavior is not affected by the category and stacklevel arguments.

The deprecation message passed to the decorator is saved in the __deprecated__ attribute on the decorated object. If applied to an overload, the decorator must be after the @~typing.overload decorator for the attribute to exist on the overload as returned by typing.get_overloads().

Added in version 3.13: See PEP 702.

Available Context Managers¶

class warnings.catch_warnings(*, record=False, module=None, action=None, category=Warning, lineno=0, append=False)¶

A context manager that copies and, upon exit, restores the warnings filter and the showwarning() function. If the record argument is False (the default) the context manager returns None on entry. If record is True, a list is returned that is progressively populated with objects as seen by a custom showwarning() function (which also suppresses output to sys.stderr). Each object in the list is guaranteed to have the following attributes:

• message: the warning message (an instance of Warning)

• category: the warning category (a subclass of Warning)

• filename: the file name where the warning occurred (str)

• lineno: the line number in the file (int)

• file: the file object used for output (if any), or None

• line: the line of source code (if available), or None

• source: the original object that generated the warning (if available), or None

Changed in version 3.6: The source attribute was added.

The type of these objects is not specified and may change; only the presence of these attributes is guaranteed.

The module argument takes a module that will be used instead of the module returned when you import warnings whose filter will be protected. This argument exists primarily for testing the warnings module itself.

If the action argument is not None, the remaining arguments are passed to simplefilter() as if it were called immediately on entering the context.

See The Warnings Filter for the meaning of the category and lineno parameters.

Note

See Concurrent safety of Context Managers for details on the concurrency-safety of the catch_warnings context manager when used in programs using multiple threads or async functions.

Changed in version 3.11: Added the action, category, lineno, and append parameters.

Concurrent safety of Context Managers¶

The behavior of catch_warnings context manager depends on the sys.flags.context_aware_warnings flag. If the flag is true, the context manager behaves in a concurrent-safe fashion and otherwise not. Concurrent-safe means that it is both thread-safe and safe to use within asyncio coroutines and tasks. Being thread-safe means that behavior is predictable in a multi-threaded program. The flag defaults to true for free-threaded builds and false otherwise.

If the context_aware_warnings flag is false, then catch_warnings will modify the global attributes of the warnings module. This is not safe if used within a concurrent program (using multiple threads or using asyncio coroutines). For example, if two or more threads use the catch_warnings class at the same time, the behavior is undefined.

If the flag is true, catch_warnings will not modify global attributes and will instead use a ContextVar to store the newly established warning filtering state. A context variable provides thread-local storage and it makes the use of catch_warnings thread-safe.

The record parameter of the context handler also behaves differently depending on the value of the flag. When record is true and the flag is false, the context manager works by replacing and then later restoring the module’s showwarning() function. That is not concurrent-safe.

When record is true and the flag is true, the showwarning() function is not replaced. Instead, the recording status is indicated by an internal property in the context variable. In this case, the showwarning() function will not be restored when exiting the context handler.

The context_aware_warnings flag can be set the -X context_aware_warnings command-line option or by the PYTHON_CONTEXT_AWARE_WARNINGS environment variable.

Note

It is likely that most programs that desire thread-safe behaviour of the warnings module will also want to set the thread_inherit_context flag to true. That flag causes threads created by threading.Thread to start with a copy of the context variables from the thread starting it. When true, the context established by catch_warnings in one thread will also apply to new threads started by it. If false, new threads will start with an empty warnings context variable, meaning that any filtering that was established by a catch_warnings context manager will no longer be active.