Implement `/foo/1/edit` route/view.
Allow setting HTTP verb by passing e.g. `/_PATCH` as the last segment of
a URL - this allows browsers to submit a `POST` request with a
pseudo-HTTP verb encoded in the URL which Jetzig can translate, i.e.
allowing forms to submit a `PATCH`.
Use `jetzig server` or `zig build -Ddebug_console=true run` to launch a
development server with the debug console enabled.
When an error is encountered a formatted stack trace is dumped to the
browser along with the current response data. Both outputs are
syntax-highlighted using Prism JS.
Add to middleware in app's `src/main.zig`:
```zig
pub const jetzig_options = struct {
pub const middleware: []const type = &.{
jetzig.middleware.AntiCsrfMiddleware,
};
};
```
CSRF token available in Zmpl templates:
```
{{context.authenticityToken()}}
```
or render a hidden form element:
```
{{context.authenticityFormElement()}}
```
The following HTML requests are rejected (403 Forbidden) if the
submitted query param does not match the value stored in the encrypted
session (added automatically when the token is generated for a template
value):
* POST
* PUT
* PATCH
* DELETE
JSON requests are not impacted - users should either disable JSON
endpoints or implement a different authentication method to protect
them.
Eradication of `data` arg to requests. We no longer need to pass this
value around as we have a) type inference, b) nested object insertion
via `put` and `append`.
Fix `joinPath` numeric type coercion
Detect empty string params and treat them as blank with expectParams()
Fix error logging/stack trace printing.
Update Zmpl - includes debug tokens + error tracing to source template
in debug builds.
Implement `request.expectParams()` to coerce params to a given struct.
`request.paramsInfo()` provides information about each param (present,
blank, failed + original values and errors where applicable).
Move all route types into a single union and remove leftover junk.
Deprecate view functions that receive a `*jetzig.Data` argument (we will
stay backward-compatible until we have a valid reason to drop support
for legacy view functions - the extra code overhead is pretty minimal).
