Mailman 3 April 2021 - Typing-sig

Next tensor typing meeting
by Matthew Rahtz 07 Apr '21

07 Apr '21

Hi everyone, We tentatively scheduled our next tensor meeting for this coming Monday the 12th of April. As it happens, though, I'm not sure we have anything significant to talk about this time, apart from some quick status updates: - PEP 646 is mostly finalised, apart from one extra section adding a bit more context around shape typing in general ( https://github.com/python/peps/pull/1904). - Guido has kindly submitted PEP 646 to the steering council ( https://github.com/python/steering-council/issues/59) - fingers crossed! Does anyone else have anything they'd like to talk about? If not, I propose we skip this month's meeting, making our next meeting Monday the 10th of May. Best, Matthew

1 1

Re: 回复：[Typing-sig] Re: 回复：[Typing-sig] Re: More forgiving TypedDict?
by Paul Bryan 07 Apr '21

07 Apr '21

+1 to the idea in principle. It can easily be implemented in a dynamic type checker (my area of interest). Some thought will need to be put into how subclassing should work (similar to the current "total" inheritance situation). Paul P.S. I was tempted to suggest some kind of wildcard key or key pattern matching, but I think that deviates from the relatively simple solution you're suggesting. On Wed, 2021-04-07 at 15:09 +0800, NGC4039 wrote: > For a third-party type checker, only the type of the value > corresponding to the defined key needs to be checked, and the value > type of the undefined key only needs to be deduced to non-existent > (just like https://www.python.org/dev/peps/pep-0655/) or Any type. > > > > ------------------ 原始邮件 ------------------ > 发件人: "Paul Bryan" <pbryan(a)anode.ca>; > 发送时间: 2021年4月7日(星期三) 下午3:03 > 收件人: "typing-sig"<typing-sig(a)python.org>; > > 主题: [Typing-sig] Re: 回复：[Typing-sig] Re: More forgiving TypedDict? > > Sorry, I didn't read the second part of your answer (constraining > types for certain keys) before I replied. > > Q3. Are you expecting that a static type checker will constrain such > types? > > Paul > > On Tue, 2021-04-06 at 23:59 -0700, Paul Bryan wrote: > > Ah, non-str values in WSGI environ. Thanks. So, rephrasing my > > second question: > > > > Q2: Since environ keys must be str, and as you don't want to > > validate the keys, what is the value of using TypedDict[...] over, > > say, dict[str, Any]? > > > > Paul > > > > On Wed, 2021-04-07 at 14:49 +0800, NGC4039 wrote: > > > I think the value type of the undefined key should be Any. > > > > > > The significance of the TypedDict that allows undefined keys is > > > that we can constrain the types of some keys and values. > > > > > > ------------------ 原始邮件 ------------------ > > > 发件人: "Paul Bryan" <pbryan(a)anode.ca>; > > > 发送时间: 2021年4月7日(星期三) 下午2:42 > > > 收件人: "NGC4039"<me(a)abersheeran.com>;"typing-sig"<typing- > > > sig(a)python.org>; > > > > > > 主题: [Typing-sig] Re: More forgiving TypedDict? > > > > > > As you point out, TypedDict was proposed to explicitly reject an > > > invalid key. > > > > > > Q1: In your view, would allowing undefined keys simply allow any > > > undefined key to be considered valid, regardless of value? > > > > > > Q2: Since environ keys and values must be str, and as you don't > > > want to validate the keys, what is the value of using > > > TypedDict[...] over, say, dict[str, str]? > > > > > > Paul > > > > > > On Wed, 2021-04-07 at 14:00 +0800, Aber wrote: > > > Hey, everyone. > > > > > > When writing a more static WSGI definition based on PEP3333, I > > > discovered such a problem. TypedDict cannot describe the HTTP_* > > > keys in the WSGI specification. I know that one of the original > > > intentions of TypedDict is to prevent typing the wrong keys. But > > > can we add an option (for example: allow_undefined=True) to allow > > > this actual problem? > > > > > > Now the definition I wrote is here > > > https://gist.github.com/abersheeran/f3723029387fe7f6584c6b157b90a > > > 50a > > > In order to allow undefined key values to be used, such as > > > `HTTP_HOST`, perhaps it can be used like the following (there is > > > currently no specification, this is just the solution I > > > envisioned). > > > > > > ```python > > > class Environ(CGIRequiredDefined, CGIOptionalDefined, > > > WSGIDefined, allow_undefined=True): > > > """ > > > WSGI Environ > > > """ > > > ``` > > > > > > Or is there a solution that I don’t know? > > > > > > --- > > > Aber > > > > > > _______________________________________________ > > > Typing-sig mailing list -- typing-sig(a)python.org > > > To unsubscribe send an email to typing-sig-leave(a)python.org > > > https://mail.python.org/mailman3/lists/typing-sig.python.org/ > > > Member address: pbryan(a)anode.ca > > > > > > > _______________________________________________ > > Typing-sig mailing list -- typing-sig(a)python.org > > To unsubscribe send an email to typing-sig-leave(a)python.org > > https://mail.python.org/mailman3/lists/typing-sig.python.org/ > > Member address: pbryan(a)anode.ca > >

1 0

Re: 回复：[Typing-sig] Re: More forgiving TypedDict?
by Paul Bryan 07 Apr '21

07 Apr '21

Ah, non-str values in WSGI environ. Thanks. So, rephrasing my second question: Q2: Since environ keys must be str, and as you don't want to validate the keys, what is the value of using TypedDict[...] over, say, dict[str, Any]? Paul On Wed, 2021-04-07 at 14:49 +0800, NGC4039 wrote: > I think the value type of the undefined key should be Any. > > The significance of the TypedDict that allows undefined keys is that > we can constrain the types of some keys and values. > > ------------------ 原始邮件 ------------------ > 发件人: "Paul Bryan" <pbryan(a)anode.ca>; > 发送时间: 2021年4月7日(星期三) 下午2:42 > 收件人: "NGC4039"<me(a)abersheeran.com>;"typing-sig"<typing- > sig(a)python.org>; > > 主题: [Typing-sig] Re: More forgiving TypedDict? > > As you point out, TypedDict was proposed to explicitly reject an > invalid key. > > Q1: In your view, would allowing undefined keys simply allow any > undefined key to be considered valid, regardless of value? > > Q2: Since environ keys and values must be str, and as you don't want > to validate the keys, what is the value of using TypedDict[...] over, > say, dict[str, str]? > > Paul > > On Wed, 2021-04-07 at 14:00 +0800, Aber wrote: > Hey, everyone. > > When writing a more static WSGI definition based on PEP3333, I > discovered such a problem. TypedDict cannot describe the HTTP_* keys > in the WSGI specification. I know that one of the original intentions > of TypedDict is to prevent typing the wrong keys. But can we add an > option (for example: allow_undefined=True) to allow this actual > problem? > > Now the definition I wrote is here > https://gist.github.com/abersheeran/f3723029387fe7f6584c6b157b90a50a > In order to allow undefined key values to be used, such as > `HTTP_HOST`, perhaps it can be used like the following (there is > currently no specification, this is just the solution I envisioned). > > ```python > class Environ(CGIRequiredDefined, CGIOptionalDefined, WSGIDefined, > allow_undefined=True): > """ > WSGI Environ > """ > ``` > > Or is there a solution that I don’t know? > > --- > Aber > > _______________________________________________ > Typing-sig mailing list -- typing-sig(a)python.org > To unsubscribe send an email to typing-sig-leave(a)python.org > https://mail.python.org/mailman3/lists/typing-sig.python.org/ > Member address: pbryan(a)anode.ca >

1 1

More forgiving TypedDict?
by Aber 07 Apr '21

07 Apr '21

Hey, everyone. When writing a more static WSGI definition based on PEP3333, I discovered such a problem. TypedDict cannot describe the HTTP_* keys in the WSGI specification. I know that one of the original intentions of TypedDict is to prevent typing the wrong keys. But can we add an option (for example: allow_undefined=True) to allow this actual problem? Now the definition I wrote is here https://gist.github.com/abersheeran/f3723029387fe7f6584c6b157b90a… In order to allow undefined key values to be used, such as `HTTP_HOST`, perhaps it can be used like the following (there is currently no specification, this is just the solution I envisioned). ```python class Environ(CGIRequiredDefined, CGIOptionalDefined, WSGIDefined, allow_undefined=True):     """     WSGI Environ     """ ``` Or is there a solution that I don’t know? --- Aber

2 1