Re: [PATCH v3 07/16] qapi/expr.py: Add casts in a few select cases

[John Snow <jsnow@redhat.com>]

On 2/24/21 7:32 AM, Markus Armbruster wrote:
> John Snow <jsnow@redhat.com> writes:
> 
>> Casts are instructions to the type checker only, they aren't "safe" and
>> should probably be avoided in general. In this case, when we perform
>> type checking on a nested structure, the type of each field does not
>> "stick".
>>
>> We don't need to assert that something is a str if we've already checked
>> that it is -- use a cast instead for these cases.
>>
>> Signed-off-by: John Snow <jsnow@redhat.com>
>> Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
>> Reviewed-by: Cleber Rosa <crosa@redhat.com>
>> ---
>>   scripts/qapi/expr.py | 8 ++++----
>>   1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/scripts/qapi/expr.py b/scripts/qapi/expr.py
>> index afa6bd07769..f45d6be1f4c 100644
>> --- a/scripts/qapi/expr.py
>> +++ b/scripts/qapi/expr.py
>> @@ -15,7 +15,7 @@
>>   # See the COPYING file in the top-level directory.
>>   
>>   import re
>> -from typing import MutableMapping, Optional
>> +from typing import MutableMapping, Optional, cast
>>   
>>   from .common import c_name
>>   from .error import QAPISemError
>> @@ -232,7 +232,7 @@ def check_enum(expr, info):
>>   
>>   
>>   def check_struct(expr, info):
>> -    name = expr['struct']
>> +    name = cast(str, expr['struct'])  # Asserted in check_exprs
>>       members = expr['data']
>>   
> 
> I believe you need this only so you can declare check_type()'s
> allow_dict to be Optional[str].  More busy-work around allow_dict...
> 
> I'm tempted to ask for allow_dict: Any and call it a day.
> 

You're right, this is because of the signature for the allow_dict 
argument. Ultimately, the pragma is a collection of strings and we need 
to prove we are looking up a string somewhere or other.

(Or tell the type checker to leave us alone.)

Unfortunately, the "check" in check_exprs falls off almost immediately. 
Working with dictly-typed objects is a little annoying for this reason.

This works for now; but finding a better way to do the pragma checks is 
probably the more correct thing. (And not something I really want to try 
and get through review until we're done typing, I think.)

>>       check_type(members, info, "'data'", allow_dict=name)
>> @@ -240,7 +240,7 @@ def check_struct(expr, info):
>>   
>>   
>>   def check_union(expr, info):
>> -    name = expr['union']
>> +    name = cast(str, expr['union'])  # Asserted in check_exprs
>>       base = expr.get('base')
>>       discriminator = expr.get('discriminator')
>>       members = expr['data']
> 
> Likwewise.
> 
>> @@ -337,7 +337,7 @@ def check_exprs(exprs):
>>           else:
>>               raise QAPISemError(info, "expression is missing metatype")
>>   
>> -        name = expr[meta]
>> +        name = cast(str, expr[meta])  # Asserted right below:
> 
> "Checked", not "asserted".
> 

Oh, yes.

>>           check_name_is_str(name, info, "'%s'" % meta)
>>           info.set_defn(meta, name)
>>           check_defn_name_str(name, info, meta)
> 
> 
> Cast before check gives me a queasy feeling.  It's lying to the type
> checker.
> 
> Hamfisted way to avoid lying:
> 
>             check_name_is_str(expr[meta], ...)
>             name = cast(str, expr[meta])
> 
> Something like
> 
>             name = check_name_is_str(name, ...)
> 
> might be neater, but then we'd have to find a better function name.
> 

OK, I'll look into re-ordering this.