[Sqlalchemy-tickets] Issue #4020: add base float coercion to float (zzzeek/sqlalchemy)

[Michael B. <iss...@bi...>]

New issue 4020: add base float coercion to float
https://bitbucket.org/zzzeek/sqlalchemy/issues/4020/add-base-float-coercion-to-float

Michael Bayer:

e.g. on MySQL, and possibly others in some cases, the driver may not be able to distinguish "float" from "numeric" and even though we know we want "float", Float currently assumes no result processing:


```
#!python

from sqlalchemy import *

e = create_engine("mysql://scott:tiger@localhost/test", echo=True)

print repr(e.scalar(select([literal(4.5, Float)])))
```

output:

    Decimal('4.5')

change would be:

```
#!diff


diff --git a/lib/sqlalchemy/sql/sqltypes.py b/lib/sqlalchemy/sql/sqltypes.py
index 7a3c505..1bf75c7 100644
--- a/lib/sqlalchemy/sql/sqltypes.py
+++ b/lib/sqlalchemy/sql/sqltypes.py
@@ -700,6 +700,8 @@ class Float(Numeric):
             return processors.to_decimal_processor_factory(
                 decimal.Decimal,
                 self._effective_decimal_return_scale)
+        elif dialect.supports_native_decimal:
+            return processors.to_float
         else:
             return None

```

the concern of course would be the extra fn call, however it does not seem to non-negligble overhead.    Oddly, the pure Python version is faster than the c processor one so we might want to consider simplfying that:


```
#!python

from sqlalchemy import processors
import timeit


def go1():
    row = []
    for fl in [3.4, 5.10, 23.235, 7.0, 18.999999]:
        row.append(fl)

proc = processors.to_float


def go2():
    row = []
    for fl in [3.4, 5.10, 23.235, 7.0, 18.999999]:
        row.append(proc(fl))

print timeit.timeit("go1", "from __main__ import go1", number=10000000)

print timeit.timeit("go2", "from __main__ import go2", number=10000000)

```

cprocessors:

    0.227154016495
    0.238603830338

non cprocessors:

    0.226067066193
    0.226836919785