Expand format_args! with more details

This commit is contained in:
hkalbasi 2023-05-16 19:12:40 +03:30
parent cbd14e9840
commit a6e5a912f9
3 changed files with 156 additions and 37 deletions

View File

@ -193,7 +193,7 @@ fn main() {
format_args!("{} {:?}", arg1(a, b, c), arg2); format_args!("{} {:?}", arg1(a, b, c), arg2);
} }
"#, "#,
expect![[r#" expect![[r##"
#[rustc_builtin_macro] #[rustc_builtin_macro]
macro_rules! format_args { macro_rules! format_args {
($fmt:expr) => ({ /* compiler built-in */ }); ($fmt:expr) => ({ /* compiler built-in */ });
@ -201,9 +201,9 @@ macro_rules! format_args {
} }
fn main() { fn main() {
$crate::fmt::Arguments::new_v1(&[], &[$crate::fmt::ArgumentV1::new(&(arg1(a, b, c)), $crate::fmt::Display::fmt), $crate::fmt::ArgumentV1::new(&(arg2), $crate::fmt::Display::fmt), ]); $crate::fmt::Arguments::new_v1(&["", " ", ], &[$crate::fmt::ArgumentV1::new(&(arg1(a, b, c)), $crate::fmt::Display::fmt), $crate::fmt::ArgumentV1::new(&(arg2), $crate::fmt::Debug::fmt), ]);
} }
"#]], "##]],
); );
} }
@ -221,7 +221,7 @@ fn main() {
format_args!("{} {:?}", a::<A,B>(), b); format_args!("{} {:?}", a::<A,B>(), b);
} }
"#, "#,
expect![[r#" expect![[r##"
#[rustc_builtin_macro] #[rustc_builtin_macro]
macro_rules! format_args { macro_rules! format_args {
($fmt:expr) => ({ /* compiler built-in */ }); ($fmt:expr) => ({ /* compiler built-in */ });
@ -229,9 +229,9 @@ macro_rules! format_args {
} }
fn main() { fn main() {
$crate::fmt::Arguments::new_v1(&[], &[$crate::fmt::ArgumentV1::new(&(a::<A, B>()), $crate::fmt::Display::fmt), $crate::fmt::ArgumentV1::new(&(b), $crate::fmt::Display::fmt), ]); $crate::fmt::Arguments::new_v1(&["", " ", ], &[$crate::fmt::ArgumentV1::new(&(a::<A, B>()), $crate::fmt::Display::fmt), $crate::fmt::ArgumentV1::new(&(b), $crate::fmt::Debug::fmt), ]);
} }
"#]], "##]],
); );
} }
@ -250,7 +250,7 @@ fn main() {
format_args!/*+errors*/("{} {:?}", a.); format_args!/*+errors*/("{} {:?}", a.);
} }
"#, "#,
expect![[r#" expect![[r##"
#[rustc_builtin_macro] #[rustc_builtin_macro]
macro_rules! format_args { macro_rules! format_args {
($fmt:expr) => ({ /* compiler built-in */ }); ($fmt:expr) => ({ /* compiler built-in */ });
@ -259,10 +259,10 @@ macro_rules! format_args {
fn main() { fn main() {
let _ = let _ =
/* parse error: expected field name or number */ /* error: no rule matches input tokens *//* parse error: expected field name or number */
$crate::fmt::Arguments::new_v1(&[], &[$crate::fmt::ArgumentV1::new(&(a.), $crate::fmt::Display::fmt), ]); $crate::fmt::Arguments::new_v1(&["", " ", ], &[$crate::fmt::ArgumentV1::new(&(a.), $crate::fmt::Display::fmt), $crate::fmt::ArgumentV1::new(&(), $crate::fmt::Debug::fmt), ]);
} }
"#]], "##]],
); );
} }

View File

@ -1,9 +1,13 @@
//! Builtin macro //! Builtin macro
use std::mem;
use ::tt::Ident;
use base_db::{AnchoredPath, Edition, FileId}; use base_db::{AnchoredPath, Edition, FileId};
use cfg::CfgExpr; use cfg::CfgExpr;
use either::Either; use either::Either;
use mbe::{parse_exprs_with_sep, parse_to_token_tree, TokenMap}; use mbe::{parse_exprs_with_sep, parse_to_token_tree, TokenMap};
use rustc_hash::FxHashMap;
use syntax::{ use syntax::{
ast::{self, AstToken}, ast::{self, AstToken},
SmolStr, SmolStr,
@ -90,11 +94,6 @@ register_builtin! {
(module_path, ModulePath) => module_path_expand, (module_path, ModulePath) => module_path_expand,
(assert, Assert) => assert_expand, (assert, Assert) => assert_expand,
(stringify, Stringify) => stringify_expand, (stringify, Stringify) => stringify_expand,
(format_args, FormatArgs) => format_args_expand,
(const_format_args, ConstFormatArgs) => format_args_expand,
// format_args_nl only differs in that it adds a newline in the end,
// so we use the same stub expansion for now
(format_args_nl, FormatArgsNl) => format_args_expand,
(llvm_asm, LlvmAsm) => asm_expand, (llvm_asm, LlvmAsm) => asm_expand,
(asm, Asm) => asm_expand, (asm, Asm) => asm_expand,
(global_asm, GlobalAsm) => global_asm_expand, (global_asm, GlobalAsm) => global_asm_expand,
@ -106,6 +105,9 @@ register_builtin! {
(trace_macros, TraceMacros) => trace_macros_expand, (trace_macros, TraceMacros) => trace_macros_expand,
EAGER: EAGER:
(format_args, FormatArgs) => format_args_expand,
(const_format_args, ConstFormatArgs) => format_args_expand,
(format_args_nl, FormatArgsNl) => format_args_nl_expand,
(compile_error, CompileError) => compile_error_expand, (compile_error, CompileError) => compile_error_expand,
(concat, Concat) => concat_expand, (concat, Concat) => concat_expand,
(concat_idents, ConcatIdents) => concat_idents_expand, (concat_idents, ConcatIdents) => concat_idents_expand,
@ -232,42 +234,159 @@ fn file_expand(
} }
fn format_args_expand( fn format_args_expand(
db: &dyn ExpandDatabase,
id: MacroCallId,
tt: &tt::Subtree,
) -> ExpandResult<ExpandedEager> {
format_args_expand_general(db, id, tt, "")
.map(|x| ExpandedEager { subtree: x, included_file: None })
}
fn format_args_nl_expand(
db: &dyn ExpandDatabase,
id: MacroCallId,
tt: &tt::Subtree,
) -> ExpandResult<ExpandedEager> {
format_args_expand_general(db, id, tt, "\\n")
.map(|x| ExpandedEager { subtree: x, included_file: None })
}
fn format_args_expand_general(
_db: &dyn ExpandDatabase, _db: &dyn ExpandDatabase,
_id: MacroCallId, _id: MacroCallId,
tt: &tt::Subtree, tt: &tt::Subtree,
end_string: &str,
) -> ExpandResult<tt::Subtree> { ) -> ExpandResult<tt::Subtree> {
// We expand `format_args!("", a1, a2)` to let args = parse_exprs_with_sep(tt, ',');
// ```
// $crate::fmt::Arguments::new_v1(&[], &[ let expand_error =
// $crate::fmt::ArgumentV1::new(&arg1,$crate::fmt::Display::fmt), ExpandResult::new(tt::Subtree::empty(), mbe::ExpandError::NoMatchingRule.into());
// $crate::fmt::ArgumentV1::new(&arg2,$crate::fmt::Display::fmt),
// ])
// ```,
// which is still not really correct, but close enough for now
let mut args = parse_exprs_with_sep(tt, ',');
if args.is_empty() { if args.is_empty() {
return ExpandResult::new(tt::Subtree::empty(), mbe::ExpandError::NoMatchingRule.into()); return expand_error;
} }
for arg in &mut args { let mut key_args = FxHashMap::default();
let mut args = args.into_iter().filter_map(|mut arg| {
// Remove `key =`. // Remove `key =`.
if matches!(arg.token_trees.get(1), Some(tt::TokenTree::Leaf(tt::Leaf::Punct(p))) if p.char == '=') if matches!(arg.token_trees.get(1), Some(tt::TokenTree::Leaf(tt::Leaf::Punct(p))) if p.char == '=')
{ {
// but not with `==` // but not with `==`
if !matches!(arg.token_trees.get(2), Some(tt::TokenTree::Leaf(tt::Leaf::Punct(p))) if p.char == '=' ) if !matches!(arg.token_trees.get(2), Some(tt::TokenTree::Leaf(tt::Leaf::Punct(p))) if p.char == '=')
{ {
arg.token_trees.drain(..2); let key = arg.token_trees.drain(..2).next().unwrap();
key_args.insert(key.to_string(), arg);
return None;
} }
} }
} Some(arg)
let _format_string = args.remove(0); }).collect::<Vec<_>>().into_iter();
let arg_tts = args.into_iter().flat_map(|arg| { // ^^^^^^^ we need this collect, to enforce the side effect of the filter_map closure (building the `key_args`)
quote! { #DOLLAR_CRATE::fmt::ArgumentV1::new(&(#arg), #DOLLAR_CRATE::fmt::Display::fmt), } let format_subtree = args.next().unwrap();
}.token_trees); let format_string = (|| {
let expanded = quote! { let token_tree = format_subtree.token_trees.get(0)?;
#DOLLAR_CRATE::fmt::Arguments::new_v1(&[], &[##arg_tts]) match token_tree {
tt::TokenTree::Leaf(l) => match l {
tt::Leaf::Literal(l) => {
let text = l.text.strip_prefix('"')?.strip_suffix('"')?;
let span = l.span;
Some((text, span))
}
_ => None,
},
tt::TokenTree::Subtree(_) => None,
}
})();
let Some((format_string, _format_string_span)) = format_string else {
return expand_error;
}; };
ExpandResult::ok(expanded) let mut format_iter = format_string.chars().peekable();
let mut parts = vec![];
let mut last_part = String::new();
let mut arg_tts = vec![];
let mut err = None;
while let Some(c) = format_iter.next() {
// Parsing the format string. See https://doc.rust-lang.org/std/fmt/index.html#syntax for the grammar and more info
match c {
'{' => {
if format_iter.peek() == Some(&'{') {
format_iter.next();
last_part.push('{');
continue;
}
let mut argument = String::new();
while ![Some(&'}'), Some(&':')].contains(&format_iter.peek()) {
argument.push(match format_iter.next() {
Some(c) => c,
None => return expand_error,
});
}
let format_spec = match format_iter.next().unwrap() {
'}' => "".to_owned(),
':' => {
let mut s = String::new();
while let Some(c) = format_iter.next() {
if c == '}' {
break;
}
s.push(c);
}
s
}
_ => unreachable!(),
};
parts.push(mem::take(&mut last_part));
let arg_tree = if argument.is_empty() {
match args.next() {
Some(x) => x,
None => {
err = Some(mbe::ExpandError::NoMatchingRule.into());
tt::Subtree::empty()
}
}
} else if let Some(tree) = key_args.get(&argument) {
tree.clone()
} else {
// FIXME: we should pick the related substring of the `_format_string_span` as the span. You
// can use `.char_indices()` instead of `.char()` for `format_iter` to find the substring interval.
let ident = Ident::new(argument, tt::TokenId::unspecified());
quote!(#ident)
};
let formatter = match &*format_spec {
"?" => quote!(#DOLLAR_CRATE::fmt::Debug::fmt),
"" => quote!(#DOLLAR_CRATE::fmt::Display::fmt),
_ => {
// FIXME: implement the rest and return expand error here
quote!(#DOLLAR_CRATE::fmt::Display::fmt)
}
};
arg_tts.push(
quote! { #DOLLAR_CRATE::fmt::ArgumentV1::new(&(#arg_tree), #formatter), },
);
}
'}' => {
if format_iter.peek() == Some(&'}') {
format_iter.next();
last_part.push('}');
} else {
return expand_error;
}
}
_ => last_part.push(c),
}
}
last_part += end_string;
if !last_part.is_empty() {
parts.push(last_part);
}
let part_tts = parts.into_iter().map(|x| {
let l = tt::Literal { span: tt::TokenId::unspecified(), text: format!("\"{}\"", x).into() };
quote!(#l ,)
});
let arg_tts = arg_tts.into_iter().flat_map(|arg| arg.token_trees);
let expanded = quote! {
#DOLLAR_CRATE::fmt::Arguments::new_v1(&[##part_tts], &[##arg_tts])
};
ExpandResult { value: expanded, err }
} }
fn asm_expand( fn asm_expand(

View File

@ -171,5 +171,5 @@ pre { color: #DCDCCC; background: #3F3F3F; font-size: 22px; padd
<span class="macro">assert</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="bool_literal macro">true</span><span class="comma macro">,</span> <span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro"> asdasd"</span><span class="comma macro">,</span> <span class="numeric_literal macro">1</span><span class="parenthesis macro">)</span><span class="semicolon">;</span> <span class="macro">assert</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="bool_literal macro">true</span><span class="comma macro">,</span> <span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro"> asdasd"</span><span class="comma macro">,</span> <span class="numeric_literal macro">1</span><span class="parenthesis macro">)</span><span class="semicolon">;</span>
<span class="macro">toho</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro">fmt"</span><span class="comma macro">,</span> <span class="numeric_literal macro">0</span><span class="parenthesis macro">)</span><span class="semicolon">;</span> <span class="macro">toho</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro">fmt"</span><span class="comma macro">,</span> <span class="numeric_literal macro">0</span><span class="parenthesis macro">)</span><span class="semicolon">;</span>
<span class="macro unsafe">asm</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="string_literal macro">"mov eax, </span><span class="format_specifier">{</span><span class="numeric_literal">0</span><span class="format_specifier">}</span><span class="string_literal macro">"</span><span class="parenthesis macro">)</span><span class="semicolon">;</span> <span class="macro unsafe">asm</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="string_literal macro">"mov eax, </span><span class="format_specifier">{</span><span class="numeric_literal">0</span><span class="format_specifier">}</span><span class="string_literal macro">"</span><span class="parenthesis macro">)</span><span class="semicolon">;</span>
<span class="macro">format_args</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="none macro">concat</span><span class="punctuation macro">!</span><span class="parenthesis macro">(</span><span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro">"</span><span class="parenthesis macro">)</span><span class="comma macro">,</span> <span class="string_literal macro">"{}"</span><span class="parenthesis macro">)</span><span class="semicolon">;</span> <span class="macro">format_args</span><span class="macro_bang">!</span><span class="parenthesis macro">(</span><span class="none macro">concat</span><span class="punctuation macro">!</span><span class="parenthesis macro">(</span><span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro">"</span><span class="parenthesis macro">)</span><span class="comma macro">,</span> <span class="string_literal macro">"</span><span class="format_specifier">{</span><span class="format_specifier">}</span><span class="string_literal macro">"</span><span class="parenthesis macro">)</span><span class="semicolon">;</span>
<span class="brace">}</span></code></pre> <span class="brace">}</span></code></pre>