use rustc::lint::*; use syntax::codemap::Span; use syntax::symbol::InternedString; use syntax::ast::*; use syntax::attr; use syntax::visit::{Visitor, walk_block, walk_pat, walk_expr}; use utils::{span_lint_and_then, in_macro, span_lint}; /// **What it does:** Checks for names that are very similar and thus confusing. /// /// **Why is this bad?** It's hard to distinguish between names that differ only /// by a single character. /// /// **Known problems:** None? /// /// **Example:** /// ```rust /// let checked_exp = something; /// let checked_expr = something_else; /// ``` declare_lint! { pub SIMILAR_NAMES, Allow, "similarly named items and bindings" } /// **What it does:** Checks for too many variables whose name consists of a /// single character. /// /// **Why is this bad?** It's hard to memorize what a variable means without a /// descriptive name. /// /// **Known problems:** None? /// /// **Example:** /// ```rust /// let (a, b, c, d, e, f, g) = (...); /// ``` declare_lint! { pub MANY_SINGLE_CHAR_NAMES, Warn, "too many single character bindings" } pub struct NonExpressiveNames { pub single_char_binding_names_threshold: u64, } impl LintPass for NonExpressiveNames { fn get_lints(&self) -> LintArray { lint_array!(SIMILAR_NAMES, MANY_SINGLE_CHAR_NAMES) } } struct ExistingName { interned: InternedString, span: Span, len: usize, whitelist: &'static [&'static str], } struct SimilarNamesLocalVisitor<'a, 'tcx: 'a> { names: Vec, cx: &'a EarlyContext<'tcx>, lint: &'a NonExpressiveNames, single_char_names: Vec, } // this list contains lists of names that are allowed to be similar // the assumption is that no name is ever contained in multiple lists. #[cfg_attr(rustfmt, rustfmt_skip)] const WHITELIST: &'static [&'static [&'static str]] = &[ &["parsed", "parser"], &["lhs", "rhs"], &["tx", "rx"], &["set", "get"], ]; struct SimilarNamesNameVisitor<'a: 'b, 'tcx: 'a, 'b>(&'b mut SimilarNamesLocalVisitor<'a, 'tcx>); impl<'a, 'tcx: 'a, 'b> Visitor<'tcx> for SimilarNamesNameVisitor<'a, 'tcx, 'b> { fn visit_pat(&mut self, pat: &'tcx Pat) { match pat.node { PatKind::Ident(_, id, _) => self.check_name(id.span, id.node.name), PatKind::Struct(_, ref fields, _) => { for field in fields { if !field.node.is_shorthand { self.visit_pat(&field.node.pat); } } }, _ => walk_pat(self, pat), } } } fn get_whitelist(interned_name: &str) -> Option<&'static [&'static str]> { for &allow in WHITELIST { if whitelisted(interned_name, allow) { return Some(allow); } } None } fn whitelisted(interned_name: &str, list: &[&str]) -> bool { list.iter().any(|&name| interned_name.starts_with(name) || interned_name.ends_with(name)) } impl<'a, 'tcx, 'b> SimilarNamesNameVisitor<'a, 'tcx, 'b> { fn check_short_name(&mut self, c: char, span: Span) { // make sure we ignore shadowing if self.0.single_char_names.contains(&c) { return; } self.0.single_char_names.push(c); if self.0.single_char_names.len() as u64 >= self.0.lint.single_char_binding_names_threshold { span_lint( self.0.cx, MANY_SINGLE_CHAR_NAMES, span, &format!("{}th binding whose name is just one char", self.0.single_char_names.len()), ); } } fn check_name(&mut self, span: Span, name: Name) { if in_macro(span) { return; } let interned_name = name.as_str(); if interned_name.chars().any(char::is_uppercase) { return; } let count = interned_name.chars().count(); if count < 3 { if count == 1 { let c = interned_name.chars().next().expect("already checked"); self.check_short_name(c, span); } return; } for existing_name in &self.0.names { if whitelisted(&interned_name, existing_name.whitelist) { continue; } let mut split_at = None; if existing_name.len > count { if existing_name.len - count != 1 || levenstein_not_1(&interned_name, &existing_name.interned) { continue; } } else if existing_name.len < count { if count - existing_name.len != 1 || levenstein_not_1(&existing_name.interned, &interned_name) { continue; } } else { let mut interned_chars = interned_name.chars(); let mut existing_chars = existing_name.interned.chars(); let first_i = interned_chars.next().expect( "we know we have at least one char", ); let first_e = existing_chars.next().expect( "we know we have at least one char", ); let eq_or_numeric = |(a, b): (char, char)| a == b || a.is_numeric() && b.is_numeric(); if eq_or_numeric((first_i, first_e)) { let last_i = interned_chars.next_back().expect( "we know we have at least two chars", ); let last_e = existing_chars.next_back().expect( "we know we have at least two chars", ); if eq_or_numeric((last_i, last_e)) { if interned_chars .zip(existing_chars) .filter(|&ie| !eq_or_numeric(ie)) .count() != 1 { continue; } } else { let second_last_i = interned_chars.next_back().expect( "we know we have at least three chars", ); let second_last_e = existing_chars.next_back().expect( "we know we have at least three chars", ); if !eq_or_numeric((second_last_i, second_last_e)) || second_last_i == '_' || !interned_chars.zip(existing_chars).all(eq_or_numeric) { // allowed similarity foo_x, foo_y // or too many chars differ (foo_x, boo_y) or (foox, booy) continue; } split_at = interned_name.char_indices().rev().next().map(|(i, _)| i); } } else { let second_i = interned_chars.next().expect( "we know we have at least two chars", ); let second_e = existing_chars.next().expect( "we know we have at least two chars", ); if !eq_or_numeric((second_i, second_e)) || second_i == '_' || !interned_chars.zip(existing_chars).all(eq_or_numeric) { // allowed similarity x_foo, y_foo // or too many chars differ (x_foo, y_boo) or (xfoo, yboo) continue; } split_at = interned_name.chars().next().map(|c| c.len_utf8()); } } span_lint_and_then( self.0.cx, SIMILAR_NAMES, span, "binding's name is too similar to existing binding", |diag| { diag.span_note(existing_name.span, "existing binding defined here"); if let Some(split) = split_at { diag.span_help( span, &format!( "separate the discriminating character by an \ underscore like: `{}_{}`", &interned_name[..split], &interned_name[split..] ), ); } }, ); return; } self.0.names.push(ExistingName { whitelist: get_whitelist(&interned_name).unwrap_or(&[]), interned: interned_name, span: span, len: count, }); } } impl<'a, 'b> SimilarNamesLocalVisitor<'a, 'b> { /// ensure scoping rules work fn apply Fn(&'c mut Self)>(&mut self, f: F) { let n = self.names.len(); let single_char_count = self.single_char_names.len(); f(self); self.names.truncate(n); self.single_char_names.truncate(single_char_count); } } impl<'a, 'tcx> Visitor<'tcx> for SimilarNamesLocalVisitor<'a, 'tcx> { fn visit_local(&mut self, local: &'tcx Local) { if let Some(ref init) = local.init { self.apply(|this| walk_expr(this, &**init)); } // add the pattern after the expression because the bindings aren't available // yet in the init // expression SimilarNamesNameVisitor(self).visit_pat(&*local.pat); } fn visit_block(&mut self, blk: &'tcx Block) { self.apply(|this| walk_block(this, blk)); } fn visit_arm(&mut self, arm: &'tcx Arm) { self.apply(|this| { // just go through the first pattern, as either all patterns // bind the same bindings or rustc would have errored much earlier SimilarNamesNameVisitor(this).visit_pat(&arm.pats[0]); this.apply(|this| walk_expr(this, &arm.body)); }); } fn visit_item(&mut self, _: &Item) { // do not recurse into inner items } } impl EarlyLintPass for NonExpressiveNames { fn check_item(&mut self, cx: &EarlyContext, item: &Item) { if let ItemKind::Fn(ref decl, _, _, _, _, ref blk) = item.node { if !attr::contains_name(&item.attrs, "test") { let mut visitor = SimilarNamesLocalVisitor { names: Vec::new(), cx: cx, lint: self, single_char_names: Vec::new(), }; // initialize with function arguments for arg in &decl.inputs { SimilarNamesNameVisitor(&mut visitor).visit_pat(&arg.pat); } // walk all other bindings walk_block(&mut visitor, blk); } } } } /// Precondition: `a_name.chars().count() < b_name.chars().count()`. fn levenstein_not_1(a_name: &str, b_name: &str) -> bool { debug_assert!(a_name.chars().count() < b_name.chars().count()); let mut a_chars = a_name.chars(); let mut b_chars = b_name.chars(); while let (Some(a), Some(b)) = (a_chars.next(), b_chars.next()) { if a == b { continue; } if let Some(b2) = b_chars.next() { // check if there's just one character inserted return a != b2 || a_chars.ne(b_chars); } else { // tuple // ntuple return true; } } // for item in items true }