Rocket/core/codegen/src/name.rs

103 lines
2.7 KiB
Rust
Raw Normal View History

use crate::http::uncased::UncasedStr;
use syn::{self, Ident, ext::IdentExt};
use proc_macro2::{Span, TokenStream};
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
/// A "name" read by codegen, which may or may not be an identifier. A `Name` is
/// typically constructed indirectly via FromMeta, or From<Ident> or directly
/// from a string via `Name::new()`. A name is tokenized as a string.
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
///
/// Some "names" in Rocket include:
/// * Dynamic parameter: `name` in `<name>`
/// * Renamed fields: `foo` in #[field(name = "foo")].
///
/// `Name` implements Hash, PartialEq, and Eq, and additionally PartialEq<S> for
/// all types `S: PartialEq<str>`. These implementations all compare the value
/// of `as_str()` only.
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
#[derive(Debug, Clone)]
pub struct Name {
value: String,
span: Span,
}
impl Name {
/// Creates a new `Name` from the string `name` and span `span`. If
/// `name` is a valid ident, the ident is stored as well.
pub fn new<S: Into<String>>(name: S, span: Span) -> Self {
Name { value: name.into(), span }
}
/// Returns the name as a string. Notably, if `self` was constructed from an
/// Ident this method returns a name *without* an `r#` prefix.
pub fn as_str(&self) -> &str {
&self.value
}
/// Like `as_str()` but into an `&UncasedStr`.
pub fn as_uncased_str(&self) -> &UncasedStr {
UncasedStr::new(self.as_str())
}
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
pub fn span(&self) -> Span {
self.span
}
}
impl devise::FromMeta for Name {
fn from_meta(meta: &devise::MetaItem) -> devise::Result<Self> {
use devise::ext::SpanDiagnosticExt;
if let syn::Lit::Str(s) = meta.lit()? {
return Ok(Name::new(s.value(), s.span()));
}
Err(meta.value_span().error("invalid value: expected string literal"))
}
}
impl quote::ToTokens for Name {
fn to_tokens(&self, tokens: &mut TokenStream) {
syn::LitStr::new(self.as_str(), self.span()).to_tokens(tokens)
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
}
}
impl From<&Ident> for Name {
fn from(ident: &Ident) -> Self {
Name::new(ident.unraw().to_string(), ident.span())
}
}
impl AsRef<str> for Name {
fn as_ref(&self) -> &str {
self.as_str()
}
}
impl std::hash::Hash for Name {
fn hash<H: std::hash::Hasher>(&self, hasher: &mut H) {
self.as_str().hash(hasher)
}
}
impl std::ops::Deref for Name {
type Target = str;
fn deref(&self) -> &Self::Target {
self.as_str()
}
}
impl Eq for Name { }
impl<S: PartialEq<str> + ?Sized> PartialEq<S> for Name {
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
fn eq(&self, other: &S) -> bool {
other == self.as_str()
Revamp codegen, fixing inconscpicuous bugs. This commit completely revamps the way that codegen handles route URI "parameters". The changes are largely internal. In summary, codegen code is better organized, better written, and less subject to error. There are three breaking changes: * `path` is now `uri` in `route` attribute: `#[route(GET, path = "..")]` becomes `#[route(GET, uri = "..")]`. * the order of execution for path and query guards relative to each-other is now unspecified * URI normalization now normalizes the query part as well. Several error messages were improved. A couple of bugs were fixed: * Prior to this commit, Rocket would optimistically try to parse every segment of a URI as an ident, in case one was needed in the future. A bug in rustc results in codegen "panicking" if the segment couldn't _lex_ as an ident. This panic didn't manifest until far after expansion, unfortunately. This wasn't a problem before as we only allowed ident-like segments (ASCII), but now that we allow any UTF-8, the bug surfaced. This was fixed by never attempting to parse non-idents as idents. * Prior to this commit, it was impossible to generate typed URIs for paths that ignored path parameters via the recently added syntax `<_>`: the macro would panic. This was fixed by, well, handling these ignored parameters. Some minor additions: * Added `RawStr::find()`, expanding its `Pattern`-based API. * Added an internal mechanism to dynamically determine if a `UriPart` is `Path` or `Query`.
2021-03-02 12:01:33 +00:00
}
}
impl std::fmt::Display for Name {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
self.as_str().fmt(f)
}
}