Avoid IEEE754 converting fraction of seconds

[timonwong]

Due to the limitations of IEEE754, we may lost precision when
converting string back to datetime objects, for example:

float('0.511581') * 1e6, gives us 0.511580, which is unexpected.

In order to address that issue, use of float is avoided.

[coveralls]

Coverage increased (+0.4%) to 88.737% when pulling c438e4f on timonwong:fix-datetime-fraction-of-seconds into 2525d0a on PyMySQL:master.

[coveralls]

Coverage increased (+0.5%) to 88.858% when pulling c438e4f on timonwong:fix-datetime-fraction-of-seconds into 2525d0a on PyMySQL:master.

[coveralls]

Coverage increased (+0.4%) to 88.742% when pulling 4229e39 on timonwong:fix-datetime-fraction-of-seconds into 2525d0a on PyMySQL:master.

[coveralls]

Coverage increased (+0.4%) to 88.742% when pulling 4229e39 on timonwong:fix-datetime-fraction-of-seconds into 2525d0a on PyMySQL:master.

[methane]

Thanks. But I feel parsing code is large enough to use regular expression.
Could you rewrite to use this?

r"(\d{1,4})-(\d{1,2})-(\d{1,2})[T ](\d{1,2}):(\d{1,2}):(\d{1,2})(?:.(\d{1,6}))?"

[coveralls]

Coverage increased (+0.3%) to 88.662% when pulling 1df94d3 on timonwong:fix-datetime-fraction-of-seconds into 2525d0a on PyMySQL:master.

[timonwong]

Thanks. But I feel parsing code is large enough to use regular expression.
Could you rewrite to use this?

r"(\d{1,4})-(\d{1,2})-(\d{1,2})T :(\d{1,2}):(\d{1,2})(?:.(\d{1,6}))?"

Yep, I rewrited both timedelta and time converters as well.

[methane]

Thank you.