python nightly failures

Development matplotlib-devel
1

The failures on python nightly are currently due to a bug in python (http://bugs.python.org/issue24176)

Tom

2

Fascinating! Can you “unpack” (heh) that error for us mere mortals? In particular:

  • never seen that “or” syntax before… Is it coercing both expressions as bool, or is it evaluating to left if bool(left) evaluates to True, else to right?

  • Why do you expect the second expression to work? Is ** supposed to have lower preference than “or”? (Which seems weird to me.)

Thanks!

Juan.

···

On Thu, May 14, 2015 at 5:08 AM, Thomas Caswell <tcaswell@…149…> wrote:

The failures on python nightly are currently due to a bug in python (http://bugs.python.org/issue24176)

Tom

One dashboard for servers and applications across Physical-Virtual-Cloud

Widest out-of-the-box monitoring support with 50+ applications

Performance metrics, stats and reports that give you Actionable Insights

Deep dive visibility with transaction tracing using APM Insight.

http://ad.doubleclick.net/ddm/clk/290420510;117567292;y

Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

3

The a or b syntax evaluates if a is ‘trueish’ and if so returns a if not returns b so c = None or {} → c == {} but c = {'a': 1} or {} → c == {‘a’: 1}

See https://docs.python.org/3.5/reference/expressions.html#grammar-token-or_test for the docs on or. and works almost the same, but returns a if a is False and b in a is True.

In the grammar for calls it should be looking for thing like “‘**’ expression” which means in the parsing anything that is part of the expression gets evaluated before the unpacking of the mapping. If you chase far enough back in the grammar an ‘or_test’ is an ‘expression’ (I may be butchering the terminology here, only just learned how lexing/parsing works a few weeks ago) so it should be fully evaluated before trying to unpack.

See https://docs.python.org/3.5/reference/expressions.html#calls for the official docs.

I suspect the source of this bug is that the grammar is getting rearranged a bit to allow for things like d = {**other_dict, ‘x’:6} and b = (*a, *c) to work as expected and something did not get changed quite right.

Tom

···

On Thu, May 14, 2015 at 5:08 AM, Thomas Caswell <tcaswell@…149…> wrote:

The failures on python nightly are currently due to a bug in python (http://bugs.python.org/issue24176)

Tom

One dashboard for servers and applications across Physical-Virtual-Cloud

Widest out-of-the-box monitoring support with 50+ applications

Performance metrics, stats and reports that give you Actionable Insights

Deep dive visibility with transaction tracing using APM Insight.

http://ad.doubleclick.net/ddm/clk/290420510;117567292;y

Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

4

Thanks Tom! Absolutely fascinating! I was trying to grok this and thinking, “but what if we want ‘or’ to return a value that will later be used as a conditional, surely it should return bool?” But of course whatever it returns will be correctly interpreted as a bool in a conditional context! Delayed/lazy bool casting, in a sense. Very clever indeed.

There’s quite a few places where this would make my code quite a bit cleaner! =)

Thanks again!

Juan.

···

On Thu, May 14, 2015 at 12:21 PM, Thomas Caswell <tcaswell@…149…> wrote:

The a or b syntax evaluates if a is ‘trueish’ and if so returns a if not returns b so c = None or {} → c == {} but c = {'a': 1} or {} → c == {‘a’: 1}

See https://docs.python.org/3.5/reference/expressions.html#grammar-token-or_test for the docs on or. and works almost the same, but returns a if a is False and b in a is True.

In the grammar for calls it should be looking for thing like “‘**’ expression” which means in the parsing anything that is part of the expression gets evaluated before the unpacking of the mapping. If you chase far enough back in the grammar an ‘or_test’ is an ‘expression’ (I may be butchering the terminology here, only just learned how lexing/parsing works a few weeks ago) so it should be fully evaluated before trying to unpack.

See https://docs.python.org/3.5/reference/expressions.html#calls for the official docs.

I suspect the source of this bug is that the grammar is getting rearranged a bit to allow for things like d = {**other_dict, ‘x’:6} and b = (*a, *c) to work as expected and something did not get changed quite right.

Tom

On Wed, May 13, 2015 at 8:33 PM Juan Nunez-Iglesias <jni.soma@…149…> wrote:

Fascinating! Can you “unpack” (heh) that error for us mere mortals? In particular:

  • never seen that “or” syntax before… Is it coercing both expressions as bool, or is it evaluating to left if bool(left) evaluates to True, else to right?
  • Why do you expect the second expression to work? Is ** supposed to have lower preference than “or”? (Which seems weird to me.)

Thanks!

Juan.

One dashboard for servers and applications across Physical-Virtual-Cloud

Widest out-of-the-box monitoring support with 50+ applications

Performance metrics, stats and reports that give you Actionable Insights

Deep dive visibility with transaction tracing using APM Insight.

http://ad.doubleclick.net/ddm/clk/290420510;117567292;y_______________________________________________

Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

On Thu, May 14, 2015 at 5:08 AM, Thomas Caswell <tcaswell@…149…> wrote:

The failures on python nightly are currently due to a bug in python (http://bugs.python.org/issue24176)

Tom

One dashboard for servers and applications across Physical-Virtual-Cloud

Widest out-of-the-box monitoring support with 50+ applications

Performance metrics, stats and reports that give you Actionable Insights

Deep dive visibility with transaction tracing using APM Insight.

http://ad.doubleclick.net/ddm/clk/290420510;117567292;y

Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

5

Thanks for tracing this Tom. I was meaning to look into it but never got round to it.

Jens

tor. 14. maj 2015 kl. 03.30 skrev Juan Nunez-Iglesias <jni.soma@…149…>:

···

On Thu, May 14, 2015 at 12:21 PM, Thomas Caswell <tcaswell@…714…> wrote:

The a or b syntax evaluates if a is ‘trueish’ and if so returns a if not returns b so c = None or {} → c == {} but c = {'a': 1} or {} → c == {‘a’: 1}

See https://docs.python.org/3.5/reference/expressions.html#grammar-token-or_test for the docs on or. and works almost the same, but returns a if a is False and b in a is True.

In the grammar for calls it should be looking for thing like “‘**’ expression” which means in the parsing anything that is part of the expression gets evaluated before the unpacking of the mapping. If you chase far enough back in the grammar an ‘or_test’ is an ‘expression’ (I may be butchering the terminology here, only just learned how lexing/parsing works a few weeks ago) so it should be fully evaluated before trying to unpack.

See https://docs.python.org/3.5/reference/expressions.html#calls for the official docs.

I suspect the source of this bug is that the grammar is getting rearranged a bit to allow for things like d = {**other_dict, ‘x’:6} and b = (*a, *c) to work as expected and something did not get changed quite right.

Tom

On Wed, May 13, 2015 at 8:33 PM Juan Nunez-Iglesias <jni.soma@…149…> wrote:

Fascinating! Can you “unpack” (heh) that error for us mere mortals? In particular:

  • never seen that “or” syntax before… Is it coercing both expressions as bool, or is it evaluating to left if bool(left) evaluates to True, else to right?
  • Why do you expect the second expression to work? Is ** supposed to have lower preference than “or”? (Which seems weird to me.)

Thanks!

Juan.

One dashboard for servers and applications across Physical-Virtual-Cloud

Widest out-of-the-box monitoring support with 50+ applications

Performance metrics, stats and reports that give you Actionable Insights

Deep dive visibility with transaction tracing using APM Insight.

http://ad.doubleclick.net/ddm/clk/290420510;117567292;y_______________________________________________

Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

On Thu, May 14, 2015 at 5:08 AM, Thomas Caswell <tcaswell@…149…> wrote:

The failures on python nightly are currently due to a bug in python (http://bugs.python.org/issue24176)

Tom

One dashboard for servers and applications across Physical-Virtual-Cloud

Widest out-of-the-box monitoring support with 50+ applications

Performance metrics, stats and reports that give you Actionable Insights

Deep dive visibility with transaction tracing using APM Insight.

http://ad.doubleclick.net/ddm/clk/290420510;117567292;y

Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel