PEP 340 Episode 2: Revenge of the With (Block)

This fornight's Python-Dev was dominated again by another nearly 400 messages on the topic of anonymous block statements. The discussion was a little more focused than the last thanks mainly to Guido's introduction of PEP 340. Discussion of this PEP resulted in a series of other PEPs, including

• PEP 342: Enhanced Iterators, which broke out into a separate PEP the parts of PEP 340 that allowed code to pass values into iterators using continue EXPR and yield-expressions.
• PEP 343: Anonymous Block Redux, a dramatically simplified version of PEP 340, which removed the looping nature of the anonymous blocks and the injection-of-exceptions semantics for generators.
• PEP 3XX: User Defined ("with") Statements, which proposed non-looping anonymous blocks accompanied by finalization semantics for iterators and generators in for loops.

Various details of each of these proposals are discussed below in the sections:

1. Enhanced Iterators
2. Separate APIs for Iterators and Anonymous Blocks
3. Looping Anonymous Blocks
4. Loop Finalization

At the time of this writing, it looked like the discussion was coming very close to a final agreement; PEP 343 and PEP 3XX both agreed upon the same semantics for the block-statement, the keyword had been narrowed down to either do or with, and Guido had agreed to add back in to PEP 343 some form of exception-injection semantics for generators.

[SJB]

Enhanced Iterators

PEP 340 incorporated a variety of orthogonal features into a single proposal. To make the PEP somewhat less monolithic, the method for passing values into an iterator was broken off into PEP 342. This method includes:

• updating the iterator protocol to use .__next__() instead of .next()
• introducing a new builtin next()
• allowing continue-statements to pass values into iterators
• allowing generators to receive values with a yield-expression

Though these features had seemed mostly uncontroversial, Guido seemed inclined to wait for a little more motivation from the co-routiney people before accepting the proposal.

Contributing threads:

• Breaking off Enhanced Iterators PEP from PEP 340

[SJB]

Separate APIs for Iterators and Anonymous Blocks

PEP 340 had originally proposed to treat the anonymous block protocol as an extension of the iterator protocol. Several problems with this approach were raised, including:

• for-loops could accidentally be used with objects requiring blocks, meaning that resources would not get cleaned up properly
• blocks could be used instead of for-loops, violating TOOWTDI

As a result, both PEP 343 and PEP 3XX propose decorators for generator functions that will wrap the generator object appropriately to match the anonymous block protocol. Generator objects without the proposed decorators would not be usable in anonymous block statements.

Contributing threads:

• PEP 340 -- loose ends
• PEP 340 -- concept clarification

[SJB]

Looping Anonymous Blocks

A few issues arose as a result of PEP 340's formulation of anonymous blocks as a variation on a loop.

Because the anonymous blocks of PEP 340 were defined in terms of while-loops, there was some discussion as to whether they should have an else clause like Python for and while loops do. There didn't seem to be one obvious interpretation of an else block though, so Guido rejected the else block proposal.

The big issue with looping anonymous blocks, however, was in the handling of break and continue statements. Many use cases for anonymous blocks did not require loops. However, because PEP 340 anonymous blocks were implemented in terms of loops, break and continue acted much like they would in a loop. This meant that in code like:

for item in items:
    with lock:
        if handle(item):
            break

the break statement would only break out of the anonymous block (the with statement) instead of breaking out of the for-loop. This pretty much shot-down PEP 340; there were too many cases where an anonymous block didn't look like a loop, and having it behave like one would have been a major stumbling block in learning the construct.

As a result, both PEP 343 and PEP 3XX were proposed as non-looping versions of PEP 340.

Contributing threads:

• PEP 340: Else clause for block statements
• PEP 340 -- loose ends
• PEP 340 -- concept clarification
• PEP 340: Breaking out.
• PEP 340: Non-looping version (aka PEP 310 redux)
• PEP 340 - Remaining issues
• PEP 340: Deterministic Finalisation (new PEP draft, either a competitor or update to PEP 340)
• Merging PEP 310 and PEP 340-redux?
• PEP 343 - Abstract Block Redux

[SJB]

Loop Finalization

Greg Ewing pointed out that a generator with a yield inside a block-statement would require additional work to guarantee its finalization. For example, if the generator:

def all_lines(filenames):
    for name in filenames:
        with open(name) as f:
            for line in f:
                yield line 

were used in code like:

for line in all_lines(filenames):
    if some_cond(line):
        break

then unless the for-loop performed some sort of finalization on the all_lines generator, the last-opened file could remain open indefinitiely.

As a result, PEP 3XX proposes that for-loops check for a __finish__() method on their iterators, and if one exists, call that method when the for-loop completes. Generators like all_lines above, that put a yield inside a block-statement, would then acquire a __finish__() method that would raise a TerminateIteration exception at the point of the last yield. The TerminateIteration exception would thus cause the block-statement to complete, guaranteeing that the generator was properly finalized.

Contributing threads:

• PEP 340 - For loop cleanup, and feature separation
• PEP 340: Deterministic Finalisation (new PEP draft, either a competitor or update to PEP 340)
• PEP 343 - Abstract Block Redux

[SJB]

Breaking out of Nested Loops

As a result of some of the issues of looping anonymous blocks, a few threads discussed options for breaking out of nested loops. These mainly worked by augmenting the break statement with another keyword (or keywords) that would indicate which loop to break out of.

One proposal suggested that break be followed with for or while to indicate which loop to break out of. But break for would only really be useful in a while-loop nested within a for-loop, and break while would only really be useful in a for-loop nested within a while-loop. That is, because loops could only be named by type, the proposal was only useful when loops of different types were mixed. This suggestion was thus discarded as not being general enough.

A few other suggestions were briefly discussed: adding labels to loops, using an integer to indicate which "stack level" to break at, and pushing breaks onto a "break buffer", but Guido killed the discussion, saying, "Stop all discussion of breaking out of multiple loops. It ain't gonna happen before my retirement."

Contributing threads:

• PEP 340: Breaking out.
• PEP 340: Deterministic Finalisation (new PEP draft, either a competitor or update to PEP 340)

[SJB]

The future of exceptions

Ka-Ping Yee suggested that instead of passing (type, value, traceback) tuples in exceptions it would be better to put the traceback in value.traceback. Guido had also suggested this (in the PEP 340 murk) but pointed out that this would not work as long as string exceptions exist (as there is nowhere to put the traceback).

Guido noted that there are no concrete plans as to when string exceptions will be deprecated and removed (other than in 3.0 at the latest); he indicated that it could be sooner, if someone wrote a PEP with a timeline (e.g. deprecated in 2.5, gone in 2.6).

Brett C. volunteered to write a PEP targetted at Python 3000 covering exception changes (base inheritance, standard attributes (e.g. .traceback), reworking the built-in exception inheritance hierarchy, and the future of bare except statements).

Contributing threads:

• Tidier Exceptions

[TAM]

Unifying try/except and try/finally

Reinhold Birkenfeld submitted a Pre-PEP to allow both except and finally clauses in try blocks. For example, a construction like:

try:
    <suite 1>
except Ex1:
    <suite 2>
<more except: clauses>
else:
    <suite 3>
finally:
    <suite 4>

would be exactly the same as the legacy:

try:
    try:
        <suite 1>
    except Ex1:
        <suite 2>
    <more except: clauses>
    else:
        <suite 3>
finally:
    <suite 4>

Guido liked this idea (so much that he wanted to accept it immediately), and recommended that it was checked in as a PEP. However, Tim Peters pointed out that this functionality was removed from Python (by Guido) way back in 0.9.6, seemingly because there was confusion about exactly when the finally clause would be called (explicit is better than implicit!). Guido clarified that control would only pass forward, and indicated that he felt that since this is now available in Java (and C#) fewer people would be confused. The main concern about this change was that, while the cost was low, it seemed to add very little value.

Contributing threads:

• Pre-PEP: Unifying try-except and try-finally

[TAM]

Decorator Library

Michele Simionato asked whether a module for commonly used decorators, or utilities to create decorators, was planned. Raymond Hettinger indicated that while this was likely in the long term, he felt that it was better if these first evolved via wikis, recipes, or mailing lists, so that a module would only be added once best practices and proven winners had emerged. In the meantime, there is both a Decorator Library wiki page and you can try out Michele's library [zip].

To assist with decorator creation, Michele would like a facility to copy a function. Phillip J. Eby noted that the informally-discussed proposal is to add a mutable __signature__ to functions to assist with signature preserving decorators. Raymond suggested a patch adding a __copy__ method to functions or a patch for the copy module, and Michele indicated that he would also like to subclass FunctionType with an user-defined __copy__ method.

Contributing threads:

• my first post: asking about a "decorator" module
• The decorator module

[TAM]

Hooking Py_FatalError

Errors that invoke Py_FatalError generally signify that the internal state of Python is in such a poor state that continuing (including raising an exception) is impossible or unwise; as a result, Py_FatalError outputs the error to stderr and calls abort(). m.u.k. would like to have a callback to hook Py_FatalError to avoid this call to abort(). The general consensus was that effort would be better directed to fixing the causes of fatal errors than hooking Py_FatalError. m.u.k.'s use case was for generating additional logging information; a callback system patch (revised by James William Pye) is available for those interested.

Contributing threads:

• Need to hook Py_FatalError

Chaining Exceptions

Ka-Ping Yee suggested adding information to exceptions when they are raised in the handler for another exception. For example:

def a():
    try:
        raise AError
    except:
        raise BError

raises an exception which is an instance of BError. This instance could have an attribute which is instance of AError, containing information about the original exception. Use cases include catching a low-level exception (e.g. socket.error) and turning it into a high-level exception (e.g. an HTTPRequestFailed exception) and handling problems in exception handling code. Guido liked the idea, and discussion fleshed out a tighter definition; however it was unclear whether adding this now was feasible - this would perhaps be best added in Python 3000.

Contributing threads:

• Chained Exceptions

[TAM]

Py_UNICODE Documentation

Nicholas Bastin started a series of threads discussing an inconsistency between the Py_UNICODE docs and the behavior on some RedHat systems. The docs say that Py_UNICODE should be an alias for wchar_t when wchar_t is available and has 16 bits, but Nick found that pyconfig.h still reports PY_UNICODE_TYPE as wchar_t, even when PY_UNICODE_SIZE is 4.

An extensive discussion between Nick, Marc-Andre Lemburg and Martin v. Löwis suggests that the possible Python-internal representations for Py_UNICODE are:

• 4-byte wchar_t encoded as UTF-32 (UCS-4)
• 2-byte wchar_t encoded as UTF-16
• unsigned short encoded as UTF-16

Python defaults to 2-byte mode, using wchar_t if available (and has 16 bits) and using unsigned short otherwise. You may end up with the 4-byte mode if TCL was built for UCS-4 (this overrides the defaults) or if you explicitly request it with --enable-unicode=ucs4. To get UCS-2 when TCL was built for UCS-4, you must explicitly request --enable-unicode=ucs2. Of course, this will mean that _tkinter can't be built anymore.

Also noted by this discussion was that even with --enable-unicode=ucs2, Python continues to support surrogate pairs in the BMP. So for example, even with a UCS-2 build, u"U00012345" encodes as a sequence of two characters; it does not produce a UnicodeError.

At the time of this posting, it did not appear that there was a documentation patch available yet.

Contributing threads:

• Py_UNICODE madness
• New Py_UNICODE doc
• Python's Unicode width default (New Py_UNICODE doc)

[SJB]

Introduction

This is a summary of traffic on the python-dev mailing list from May 01, 2005 through May 15, 2005.

It is intended to inform the wider Python community of on-going developments on the list on a semi-monthly basis. An archive of previous summaries is available online.

An RSS feed of the titles of the summaries is available. You can also watch comp.lang.python or comp.lang.python.announce for new summaries (or through their email gateways of python-list or python-announce, respectively, as found at http://mail.python.org).

This is the third summary written by the python-dev summary cabal of Steve Bethard, Tim Lesher, and Tony Meyer.

To contact us, please send email:

• Steve Bethard (steven.bethard at gmail.com)
• Tim Lesher (tlesher at gmail.com)
• Tony Meyer (tony.meyer at gmail.com)

Do not post to comp.lang.python if you wish to reach us.

The Python Software Foundation is the non-profit organization that holds the intellectual property for Python. It also tries to advance the development and use of Python. If you find the python-dev Summary helpful please consider making a donation. You can make a donation at http://python.org/psf/donations.html . Every penny helps so even a small donation with a credit card, check, or by PayPal helps.

Commenting on Topics

To comment on anything mentioned here, just post to comp.lang.python (or email python-list at python.org which is a gateway to the newsgroup) with a subject line mentioning what you are discussing. All python-dev members are interested in seeing ideas discussed by the community, so don't hesitate to take a stance on something. And if all of this really interests you then get involved and join python-dev!

How to Read the Summaries

The in-development version of the documentation for Python can be found at http://www.python.org/dev/doc/devel/ and should be used when looking up any documentation for new code; otherwise use the current documentation as found at http://docs.python.org/ . PEPs (Python Enhancement Proposals) are located at http://www.python.org/peps/ . To view files in the Python CVS online, go to http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/python/ . Reported bugs and suggested patches can be found at the SourceForge project page.

Please note that this summary is written using reStructuredText. Any unfamiliar punctuation is probably markup for reST (otherwise it is probably regular expression syntax or a typo =); you can safely ignore it. I do suggest learning reST, though; it's simple and is accepted for PEP markup and can be turned into many different formats like HTML and LaTeX. Unfortunately, even though reST is standardized, the wonders of programs that like to reformat text do not allow me to guarantee you will be able to run the text version of this summary through Docutils as-is unless it is from the original text file.