tutu/docs/devnotes.md

87 lines
4.3 KiB
Markdown
Raw Normal View History

2024-08-22 11:46:00 +00:00
# Dev Notes
This file documents some advices working on the app.
## Debugging for iOS
You can debug on the Safari on iOS only if you have mac (and run macOS). The certain workarounds don't work well. Here are some tips:
- [GNOME Web (epiphany)](https://apps.gnome.org/Epiphany/) uses webkit2gtk, the same engine from Safari. Most of bugs can be reproduced on it.
- For visual bugs: on you iDevice, redirect the localhost.direct to your dev computer. Now you have the hot reload on you iDevice.
- You can use network debugging apps like "Shadowrocket" to do such thing.
## The components don't react to the change as I setting the store, until the page reloaded
The `WritableAtom<unknwon>.set` might do an equals check. You must set a different object to ensure the atom sending a notify.
The code below may not notify the change:
```ts
export function updateAcctInf(idx: number) {
const o = $accounts.get();
// ...
o[idx].inf = inf;
$accounts.set(o);
}
```
Instead, set a new object:
```ts
export function updateAcctInf(idx: number) {
const o = $accounts.get();
// ...
o[idx] = Object.assign({}, o[idx], { inf });
$accounts.set(Array.from(o));
}
```
Ja, the code is weird, but that's the best we know. Anyway, you need new object on the path of your changed value.
## `transition: *-block or *-inline` does not work on WebKit
Idk why, but transition on logical directions may not work on WebKit - sometimes they work.
Use physical directions to avoid trouble, like "margin-top, margin-bottom".
## Safe area insets
2024-11-19 09:58:01 +00:00
For isolating control of the UI effect, we already setup css variables `--safe-area-inset-*`. In components, you should use the variables unless you have reasons to use `env()`.
2024-11-19 09:58:01 +00:00
Using `--safe-area-inset-*`, you can control the global value in settings (under dev mode).
2024-11-22 09:16:56 +00:00
## Module Isolation
> Write the code that can be easily removed.
To limit the code impact, we organize the code based on **"topic modules"** (modules in short). Each module focus on a specific topic described by the name. Like the "accounts" contains the code about the accounts, "masto" contains the code about the masto (a library used to access mastodon) helpers.
> Sidenote: This also helps easing "the landing problem". If you need something about accounts, no longer "common/accounts" and "hooks/accounts" and "helpers/accounts" and "components/accounts". Someone says this is clean - is it even if you need to jump between 6 directories for how one simple feature works?
> And you no longer needs to think about "where to place this file (between six directories, usually)". People often optimize their code structure too early - just like how they treat the runtime performance.
> The worse is, it's very hard to solve this problem later, because you had sent your code to different places.
There are **two special modules** in this project:
One is the *platform*. This module provides foundation of this app: deals with the host platform (like SizedTextarea - auto resized textarea), provides custom platform feature (like StackedRouter - provides mobile-native navigation experience).
The another is the *material*. This module provides Material styling toolkit, the stylesheets, MUI Theme, constants and components.
They (and only them) can be accessed by special aliases: `~{module name}`, like the `~platform`.
We discourage cross referencings between two topics. Reuse is not better than duplication. Cross referencing is still possible if required.
When a tool, a file or a component is required every-elsewhere, **promoting** is required to reduce the cross referencing. Thanksfully, it's usually automated process for moving files.
But, sometimes you need a redesigned (sometimes better) tool for the generic usage. Follow the idea:
- Move slowly or crash. Only make the change if it's required.
- Try to make the original part depends on your new tool, and keep the original for awhile.
- Mark deprecated only if you think the original won't worth an existence. Reasons:
- Migrate to the new code only needs minor change.
- The original code has critical problems, like performance or compatibility.
- Make notes. Communication is important, even with the future you.
- *Why* this move is decided?
- *What* this new tool does?
- *How* this tool works?
- Clean up code regularly. Don't keep the unused code forever.