Add int support #3
Conversation
| defp encode_int(data), do: <<data::signed-256>> | ||
|
|
||
| defp signed_overflow?(n, max_bits) do | ||
| n < (:math.pow(2, max_bits-1) * -1) + 1 || n > (:math.pow(2, max_bits-1)) - 1 |
There was a problem hiding this comment.
@ayrat555, this will mathematically check if the given data causes an integer overflow with the given number of bits. I see the overflow checking is implemented differently in the uint encoding (by converting binaries in Elixir and counting bytes). Do you know why it was implemented that way and not with a formula (probably something like n > (:math.pow(2, max_bits)) - 1)?
I am asking because I am a little bit worried that I might not be seeing something and my solution is actually wrong.
There was a problem hiding this comment.
Are you sure :math.pow is faster? Usually at the C level, powers of 2 are done with bit shifts. It may be that counting bytes in a binary is faster in Elixir/Erlang than :math:pow. I'm not sure if :math.pow is optimized for base 2 in the C code of the BIF.
There was a problem hiding this comment.
I'm not sure if it's faster, but I couldn't use the same approach used for uint because it relies on :binary.encode_unsigned/1, which doesn't work for negative values. This is the simplest approach I could come up with. I didn't even think too much about the (possible) performance issue because I thought my solution could be even faster than the uint one because I was just making a simple calculation, while the other one was converting to binary, counting bytes and comparing with another number (that also had to be calculated). But that was just a feeling, so thanks for bringing that up. Do you think we can improve the performance?
You can see the uint encoding and overflow check here: https://github.com/Lucasnar/ex_abi/blob/92e14bf674e41672d6958e2e85834867aee14886/lib/abi/type_encoder.ex#L252
There was a problem hiding this comment.
I would recommend running benchee tests to be sure, but that can be done in other work.
|
@lucasnar can you bump the library version? |
Resolves #1 (originally: blockscout/blockscout#363).
Description
The
inttype in Solidity differs fromuintin which it includes negative numbers in its range of possible values. For instance, an 8-bitintcan contain any number from -127 to 127, while an 8-bituintcan contain any number from 0 to 255. Also, if we try to define an 8-bit integer outside of those ranges, we will get an overflow (and possibly an error or unexpected behaviour from the machine that is running the code). Related resource on unsigned and signed integers: http://kias.dyndns.org/comath/13.html.While this lib could handle
uinttypes, it had no support forinttypes. This PR adds support for them, by enabling both encoding and decoding ofinttypes in theABI.TypeEncoderandABI.TypeDecoder.