Roman Sharkov

@pavelnikolov breaking the API is never a good option. However, `ValidationOpt` feels unnecessarily complex. I can't imagine what kind of validation options there'd be other than parameters. Why not just...

We're currently using [neelance/graphql-go](https://github.com/neelance/graphql-go) ontop of a WebSocket server. Basically a GraphQL server doesn't (or shouldn't) care about the transport layer. It receives a query, returns a result, that's it......

Can you please provide the exact code for reproduction?

Alternatively, the error reporting can be changed to avoid needing `Src` in the first place. The downside would be slightly less convenient error messages: ``` error at index 6: illegal...

Support for [io.Reader](https://pkg.go.dev/io#Reader) is already in the works. Since it requires extensive changes to the API and the algorithm I decided to include it in later feature bumps above [v2.0.0](https://github.com/romshark/jscan/releases/tag/v2.0.0)....

QuickStreams is **not a product**, it's a proof of concept implementation of asynchronous programming in streams (as the README says). Work on this particular repository was suspended for quite a...

> Should those be in docstrings or distributed across the code by itself? In the doc comments of the individual functions/methods, like: ```go // ScanOne calls fn for every encountered...

To understand why I don't return `error`, as is common in Go, consider the following benchmark (https://go.dev/play/p/_iAcdhVU8m7): ``` === RUN TestSizeOf bench_test.go:10: size of Error[string]: 32 bench_test.go:11: size of Error[[]byte]:...

### Using `jscan.Error[T]` as `error` > but in this case, the user still needs to make an effort to avoid `fmt.Sprintf` or anything like that. Not much effort required since...

Eliminating unnecessary jump in number parser: ``` goos: darwin goarch: arm64 pkg: github.com/romshark/jscan/v2 │ large_old.txt │ large_new.txt │ │ sec/op │ sec/op vs base │ Valid/large_26m_____________-10 11.18m ± 0% 11.17m...