Remove Runtime Exceptions

[johannes-wolf]

It is still easy to accidentally "crash" the WASM runtime with queries – so let's replace runtime exceptions with tl::expected.

Missing:

• Division by zero still throws
• The model still throws (4 exceptions left)
• Function arg-parsing (at runtime) still throws

On main:

benchmark name                       samples       iterations    est run time
                                     mean          low mean      high mean
                                     std dev       low std dev   high std dev
-------------------------------------------------------------------------------
Query typeof Recursive                         100             1     1.73898 s 
                                        17.6426 ms    17.5999 ms     17.695 ms 
                                        241.428 us    204.215 us    290.245 us 
                                                                               
Query field id Recursive                       100             1     1.36372 s 
                                        13.9724 ms    13.8896 ms    14.0689 ms 
                                        456.841 us    402.926 us    551.696 us 
                                                                               
Query field id                                 100             1    28.5413 ms 
                                        286.066 us    284.083 us     289.23 us 
                                        12.5867 us     8.7692 us    17.7284 us

This branch:

benchmark name                       samples       iterations    est run time  
                                     mean          low mean      high mean     
                                     std dev       low std dev   high std dev  
-------------------------------------------------------------------------------
Query typeof Recursive                         100             1     1.56551 s 
                                        15.1817 ms    15.1624 ms    15.2042 ms 
                                        106.495 us    89.1097 us    127.841 us 

Query field id Recursive                       100             1    940.211 ms 
                                         9.4754 ms     9.4566 ms    9.49996 ms 
                                        109.514 us    87.3406 us    139.911 us 

Query field id                                 100             1    23.6667 ms 
                                        233.751 us    231.936 us    240.035 us 
                                        15.4148 us    4.62786 us    34.9265 us 

Query field keys recursive                     100             1     1.26274 s 
                                        12.5335 ms    12.5022 ms    12.5729 ms 
                                        178.346 us    147.985 us    225.614 us

---

Note

Switches the engine from throwing exceptions to returning errors via tl::expected across expressions, functions, model/value, I/O, and tooling, updating APIs and tests accordingly.

• Error handling (core):
  • Replace runtime exceptions with tl::expected throughout (Expr::eval/ieval, Function::eval, ResultFn, MetaType, operators dispatch, etc.).
  • Expand simfil::Error::Type and add equality for Error/SourceLocation.
• Expressions/Operators:
  • All eval paths now return tl::expected<Result, Error>; propagate errors and remove exception-based control flow (e.g., division by zero -> Error::DivisionByZero).
  • Add move-optimized eval overloads; debug paths adjusted.
• Functions:
  • Function interface returns expected; argument parsing reworked to validate/count/type-check via ArgParser without throwing.
• Value/Model:
  • Refactor Value to carry ModelNode::Ptr in its variant; add helpers (nodePtr, node, asBool).
  • Model/ModelPool APIs (resolve, root, validate, read/write, setStrings) now return expected; error messages standardized.
  • StringPool::emplace/read/write return expected; handle pool overflow as Error.
• JSON/Serialization:
  • json::parse and internal build use expected; Bitsery traits handle arena access errors.
• Arena:
  • ArrayArena::at now returns expected<reference_wrapper<...>, Error>; iterator deref asserts; out-of-range returns Error::IndexOutOfRange.
• Overlay/Completion/Parser:
  • Adjust to new eval and model APIs; non-cloneable completion exprs unchanged but error-aware.
• Examples/REPL:
  • Update to unwrap expected (root/eval/parse) and print errors instead of throwing.
• Tests:
  • Update expectations to check expected results and new error cases; add performance/diagnostics tweaks.

^{Written by Cursor Bugbot for commit f1a26a5. This will update automatically on new commits. Configure here.}

[github-actions]

Test Results

 1 files  ± 0   1 suites  ±0   6m 53s ⏱️ + 4m 2s
86 tests +31  86 ✅ +31  0 💤 ±0  0 ❌ ±0 
91 runs  +31  91 ✅ +31  0 💤 ±0  0 ❌ ±0 

Results for commit 90615c1. ± Comparison against base commit 1a2b998.

This pull request removes 2 and adds 33 tests. Note that renamed tests count towards both.

CompleteFieldOrString ‑ CompleteFieldOrString
CompleteSorted ‑ CompleteSorted

Binary arithmetic operators ‑ Binary arithmetic operators
Binary comparison operators ‑ Binary comparison operators
Complete And/Or ‑ Complete And/Or
Complete Field with Special Characters ‑ Complete Field with Special Characters
Complete Function ‑ Complete Function
Complete SmartCas ‑ Complete SmartCas
Complete Wildcard Comparison ‑ Complete Wildcard Comparison
Complete Wildcard Hint ‑ Complete Wildcard Hint
Complete in Expression ‑ Complete in Expression
Complete in unclosed expression ‑ Complete in unclosed expression
…

♻️ This comment has been updated with latest results.

[johannes-wolf]

Outside of constructors there are 4 calls to raise<...> left in the codebase with this PR.
(Plus some unreachable throws for unimplemented virtual functions that must never get called.)

[josephbirkner]

Looks good! We could probably improve performance even more by moving from result callbacks to coroutines. Should I create a ticket for this?

[cursor]

This PR is being reviewed by Cursor Bugbot

Details

Your team is on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle for each member of your team.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

[josephbirkner]

Looks like a release 🥳

Might we think about doing this?

if (!result) {      // GCOVR_EXCL_START
  return Err{ENull};
}                   // GCOVR_EXCL_STOP

[johannes-wolf]

I've added the gcovr option --exclude-lines-by-pattern TRY_EXPECTED and replaced most of the if (!res) ... blocks with the macro. EDIT: Does not seem to work 🙃.

See: https://gcovr.com/en/stable/manpage.html#cmdoption-gcovr-exclude-lines-by-pattern

[sonarqubecloud]

Quality Gate passed

Issues
39 New issues
4 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
2.0% Duplication on New Code

See analysis details on SonarQube Cloud

[github-actions]

Package	Line Rate	Branch Rate	Health
include.simfil	27%	11%	❌
include.simfil.model	91%	57%	✔
src	74%	46%	➖
src.model	80%	45%	✔
Summary	45% (5783 / 12752)	27% (3880 / 14313)	❌

[josephbirkner]

🍰