Changed Avro type for Python int to long

[panasenco]

... and Avro type for Python float to double.

Reasoning:

• "Integers have unlimited precision." [1]
• "Floating point numbers are usually implemented using double in C;" [1] and "almost all platforms map Python floats to IEEE-754 “double precision”." [2]

[panasenco]

Haven't actually ensured unit tests pass, brb...

[codecov-io]

Codecov Report

Merging #83 (81d1ff4) into master (2ee644f) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master      #83   +/-   ##
=======================================
  Coverage   99.50%   99.50%           
=======================================
  Files           7        7           
  Lines         607      608    +1     
  Branches       93       93           
=======================================
+ Hits          604      605    +1     
  Misses          1        1           
  Partials        2        2           

Impacted Files	Coverage Δ
dataclasses_avroschema/fields.py	99.45% <100.00%> (+<0.01%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2ee644f...81d1ff4. Read the comment docs.

[marcosschroh]

Hi @panasenco ,

Thanks for the PR. Make sense your changes. I am wondering whether this can break the current process of the users. I Think won't be a problem as avro considers:

avro type	json type	example
int,long	integer	1.
float, double	number	1.1

With this changes the resulting schema will replace int by long and float by double and I think everything should work. Any ideas?

[panasenco]

Hey Marcos,

Honestly, the change will probably be breaking some existing schemas if people upgrade without reading the changelog.

However, I think that even those people will agree the change is necessary. I was testing the schema with an integer like 12345678950 and was getting an overflow when using Kafka Connect to write to a SQL table. When people work with Python they expect practically unlimited integers and double-precision floats, because that's what Python has.

I think the benefits far outweigh the risks on this one. :)

Aram

[panasenco]

That said, I went ahead and added some notices in big letters for people checking out the GitHub page and the documentation so that they are unlikely to miss it before upgrading.

[marcosschroh]

Cool! Let's merge it! Thanks!