![\"Sql](\"/imgs/sqlflow.png\")
If you have problem with uperfect monitor and you are using with MacBook then this is for you.
\nThe coil whine is present when the display is powered only from MacBook.
\nUperfect monitors support power pass through which actually removes the coil whine.
\nSo plug MacBook power source into the monitor and connect monitor to the MacBook. This way the mac gets all the power and the monitor won't whine.
\n", "url": "https://rybarix.com/tils/uperfect-coil-whine.html", "title": "Uperfect external monitor with coil whine", "date_modified": "2026-03-17T00:00:00.000Z", "tags": [ "TIL" ] }, { "id": "https://rybarix.com/tils/swift-state.html", "content_html": "@StateMarks view's property as reactive. Like React's useState() or Vue's ref().
Triggers rendering when changed.
@BindingSame concept as v-model in Vue. Marks view's property for 2-way binding.
Modifying the property inside view will propagate changes outside.
ObservableObject Protocol and @ObservedObjectclass OrderModelController: ObservableObject {\n // @Published notifies watchers when order is modified\n @Published var order: Order\n}\n\n\nclass SomeView: View {\n // This may cause issues during re-render when the @ObservedObject is owned by view and is not injected\n @ObservedObject var orderController = OrderModelController.load()\n // ^ Use @StateObject in this case.\n}\nIn general you want something like this:
\n// SomeView.swift\nclass SomeView: View {\n @ObservedObject var injectedState: MyState\n init(myState: MyState) {\n self.injectedState = myState\n }\n}\n\n// SomeModel.swift\nclass SomeModel: ObservableObject {\n @Published var stateVariable: MyState\n\n init() {\n // initialize model\n }\n}\n\n// We are sharing the state from top to the bottom.\n// This prop drilling can be mitigated by\n// Entrypoint\n// MyApp.swift\n@main\nstruct MyApp: App {\n @StateObject private var myState: MyState\n\n init() {\n self._myState = StateObject(wrappedValue: MyState())\n }\n\n var body: some View {\n // View here\n }\n}\nHow to make SQL more readable?
\nIs it even good question when it is closer to english than any other programming language?
\nMy bet (and hope) is that SQL will be more common knowledge as spreadsheet formulas are.
\nIt's expressive and quite powerful and get works done. I would say that it is much more easier to write than it is to read.
\nComposing queries in your head will make sense to you as you go step by step. Reading others larger SQL statements will be challenging as you try understand how the other person went to solve a problem.
\nQuite a lot of "Explain SQL AI" solutions popped up lately. And I get it that there is a need to break down the SQL to natural language. But the question is: Can you trust it? Nope.
\nSo, we are not leaving the code anytime soon. And that's good thing! How to present SQL to novices or people new to the field?
\nIf only there was a SQL to Spreadsheet formulas translation. But there isn't and it wouldn't make things easier based on my experience with large nested formulas from complicated spreadsheets.
\nCurrently, my best solution is to break the SQL query into series of simpler steps and display it as a visual graph.\nWhere you start at the left and work towards to the right. It's not perfect but it is deterministic and doesn't hide any details. The vision is to invite users to learn not isolate them. At the end of the day, who would take responsibility for outputs that they can't verify or trust?
\n
A nice side-effect of this approach is the ability to run query partially and understand the temporary results. It's not perfect but it's much better than big SQL snippet.
\n", "url": "https://rybarix.com/tils/simpler-sql.html", "title": "Simpler SQL", "date_modified": "2026-02-17T00:00:00.000Z", "tags": [ "TIL" ] }, { "id": "https://rybarix.com/tils/rs-enum-mem.html", "content_html": "Something not immediately obvious but obvious after thinking about it for a minute is memory size of tagged unions.
\nEach Token will take up 40B + size of enum discriminant - 1B (tag) which will be aligned to 8B because entire structure needs to be aligned by the highest data type size (usize or String (ptr inside is 8B)) which is 8B.
\nstruct Span {\n line: usize, // 8 bytes\n column: usize, // 8 bytes\n}\n\nenum Token {\n Identifier {\n name: String, // 24 bytes\n location: Span // 16 bytes\n },\n Number {\n value: i32, // 4 bytes\n raw: String // 24 bytes\n },\n If(Span), // 16 bytes\n}\nI was thinking about this for more than one year now.
\nIt is time to give it a shot and create brand new scripting language that is secure by default.
\nWhat security means in this context?
\nThe main goal is to have runtime and allow user scripting without relying on OS APIs.
\nEvery single general-purpose programming language assume that code is written by trusted programmer.
\nIf it was easy to turn any programming language into safe one, this page wouldn't exist.
\nWhy not WASM?
\nWASM is promising runtime and combined with WASI, it should be useful for sandboxing other programming languages.
\nBut using WASM prevents to see what would be possible if langauge itself was built with sandboxing in mind from day 1.
\n", "url": "https://rybarix.com/dev/lang/init.html", "title": "Sandbox-first language", "date_modified": "2026-02-04T00:00:00.000Z", "tags": [ "Dev Log" ] }, { "id": "https://rybarix.com/tils/ai-anthropo.html", "content_html": "You can't talk to the AI without treating it like a human.
\nThey say: "Treat it like a tool".
\nBut how can you instruct the tool without "talking" it to it like a human?!
\n", "url": "https://rybarix.com/tils/ai-anthropo.html", "title": "Can you treat AI as a tool?", "date_modified": "2026-01-13T00:00:00.000Z", "tags": [ "TIL" ] } ] }