Const fields

[g-r-a-n-t]

What is wrong?

There is no way for a user to specify that a field is constant.

This would allow for significant gas saving, since we can store constant values in bytecode, instead of storage.

How can it be fixed

Do something similar to the work done on static strings. The syntax should look something like this: const MINIMUM_LIQUIDITY: u256 = 1000

[rahulbansal16]

I am trying to solve this and I was going through the codebase to understand the implementation. I have figured out the following steps. I have a few doubts. If you can clear those it will help me speed up.

1. Add a Token called CONST in the types.rs.
2. Create the logic inside the tokenizer.rs to parse the const token. I was trying to understand where should I put the statement to parse the const token. I was looking for the All the token list but I was unable to find one. Where are token like contract, type, return are tokenised?
3. Write the logic to parse a token for the const inside the parser.rs. I am unable to understand the purpose of this file? Aren't the tokenizer.rs going to create the tokens from the source code? I see the methods like string_token but they are not called anywhere except the test directory. What is the purpose of those methods?
4. Define a grammar rule inside the Fe.grammar file which will be used to create the const expression. I feel the rule is already added inside the file as contract_field_qual so this step is not required.
5. I am still figuring out the steps after that.

[g-r-a-n-t]

Hi @rahulbansal16, thank you very much for reaching out.

So yeah, as you noted, const is already included in the grammar file and is also included in the parser (see this test).

The grammar is somewhat incorrect, though, in that we can't assign a value to a field, which is of course what we need to do in the case of a constant field.

Currently, we have:

contract_field: [contract_field_qual] NAME ':' type_desc NEWLINE

when we should have:

contract_field: [contract_field_qual] NAME ':' type_desc ['=' expr] NEWLINE

Making this change to the grammar file, updating the parser, and then creating a test to make sure a statement like const foo: u256 = 42 is parsed would be the first step in adding support for constant fields. If you're still interested in helping out, I can provide more guidance. It would definitely be a big help.

[rahulbansal16]

Thanks @g-r-a-n-t I am making changes to add the feature. The PR is #196
It's a WIP. Do take a look if I am taking the correct approach. I have split the grammar rule into two rules.

I was trying to run the test locally on the master branch and I got the following errors for the tests for the test_parser.rs file.

---- test_target stdout ----
thread 'test_target' panicked at 'Test example has wrong format fixtures/parsers/target.ron: "expected 2 parts, got 1"', parser\tests\test_parsers.rs:284:1

I am using a windows system and I am not doing a full build.
I used the following command to run the build and test.
cargo +nightly build
cargo +nightly test --workspace
I used nighty because the rust compiler was asking me to use it.

[cburgdorf]

I am using a windows system and I am not doing a full build.

Sorry to hijack the issue but..Any chance you'd be interested to add CI coverage + Release builds for Windows? I tried to add Windows to our CI suite a while back in #64 but failed to get it to work. None of us is on a windows machine so having someone familiar with windows looking into it would be very much welcome :)

[rahulbansal16]

No worries @cburgdorf, I can start looking into it after I am done with this issue.

[rahulbansal16]

@cburgdorf I am able to download the boost and updating the path. I am getting link errors. I tried the multiple version of the boost and the error is for the boost version 1.69.0 refer to the build logs for more details.

msvcprtd.lib(MSVCP140D.dll) : error LNK2005: "public: __cdecl std::_Locinfo::~_Locinfo(void)" (??1_Locinfo@std@@qeaa@XZ) already defined in libboost_program_options-vc141-mt-sgd-x64-1_69.lib(options_description.obj) [D:\a\fe\fe\target\debug\build\solc-6fd6a06f62b98451\out\build\solc\solc.vcxproj]

libboost_program_options-vc141-mt-sgd-x64-1_69.lib(value_semantic.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MTd_StaticDebug' doesn't match value 'MDd_DynamicDebug' in CommandLineInterface.obj [D:\a\fe\fe\target\debug\build\solc-6fd6a06f62b98451\out\build\solc\solc.vcxproj]

Do you have any idea about fixing them?

[rahulbansal16]

I was also trying with the boost version of 1.73.0 and the error in the build logs is

but it set boost_filesystem_FOUND to FALSE so package "boost_filesystem" is
considered to be NOT FOUND. Reason given by package:
No suitable build variant has been found. * libboost_filesystem-vc141-mt-gd-x64-1_73.lib (vc141, detected vc142, set
Boost_COMPILER to override)
* libboost_filesystem-vc141-mt-s-x32-1_73.lib (32 bit, need 64)

I have few approaches that I am trying. Here Have you encountered these errors before?

[cburgdorf]

@rahulbansal16 Let's discuss this over in #198 I have a rough idea regarding the boost issue.