codelens_engine/

auto_import.rs

//! Automatic import suggestion and insertion.
//!
//! Detects unresolved symbols in a file and suggests imports from the project's symbol index.
//! Generates language-appropriate import statements and inserts at the correct position.
//!
//! **Internal API.** `add_import` writes through
//! `apply_full_write_with_evidence` and bypasses ADR-0009 role
//! gates, audit sinks, and cache invalidation when called outside
//! `codelens-mcp` dispatch. Route mutations through MCP — see the
//! crate-level docs in `lib.rs`.

use crate::import_graph::extract_imports_for_file;
use crate::project::ProjectRoot;
use crate::symbols::{SymbolIndex, SymbolInfo, get_symbols_overview, language_for_path};
use anyhow::Result;
use regex::Regex;
use serde::Serialize;
use std::collections::{HashMap, HashSet};
use std::fs;
use std::hash::{Hash, Hasher};
use std::path::Path;
use std::sync::{LazyLock, Mutex};
use tree_sitter::{Node, Parser};

static TYPE_CANDIDATE_RE: LazyLock<Regex> =
    LazyLock::new(|| Regex::new(r"\b([A-Z][a-zA-Z0-9_]*)\b").unwrap());

const IMPORT_CACHE_CAPACITY: usize = 64;

static IMPORT_ANALYSIS_CACHE: LazyLock<Mutex<HashMap<u64, MissingImportAnalysis>>> =
    LazyLock::new(|| Mutex::new(HashMap::new()));

fn content_cache_key(file_path: &str, content: &str) -> u64 {
    let mut hasher = std::hash::DefaultHasher::new();
    file_path.hash(&mut hasher);
    content.hash(&mut hasher);
    hasher.finish()
}

#[derive(Debug, Clone, Serialize)]
pub struct ImportSuggestion {
    pub symbol_name: String,
    pub source_file: String,
    pub import_statement: String,
    pub insert_line: usize,
    pub confidence: f64,
}

#[derive(Debug, Clone, Serialize)]
pub struct MissingImportAnalysis {
    pub file_path: String,
    pub unresolved_symbols: Vec<String>,
    pub suggestions: Vec<ImportSuggestion>,
}

/// Analyze a file for potentially unresolved symbols and suggest imports.
/// Results are cached by (file_path, content_hash) to avoid redundant parsing and lookups.
pub fn analyze_missing_imports(
    project: &ProjectRoot,
    file_path: &str,
) -> Result<MissingImportAnalysis> {
    let resolved = project.resolve(file_path)?;
    let source = fs::read_to_string(&resolved)?;
    let cache_key = content_cache_key(file_path, &source);

    // Return cached result if file content unchanged
    if let Ok(cache) = IMPORT_ANALYSIS_CACHE.lock()
        && let Some(cached) = cache.get(&cache_key)
    {
        return Ok(cached.clone());
    }

    let ext = resolved
        .extension()
        .and_then(|e| e.to_str())
        .unwrap_or("")
        .to_ascii_lowercase();

    // Step 1: Extract type identifiers via tree-sitter (not regex)
    let used_types = collect_type_candidates_ast(&resolved, &source)?;

    // Step 2: Collect locally defined symbols
    let local_symbols: HashSet<String> = get_symbols_overview(project, file_path, 0)?
        .into_iter()
        .flat_map(flatten_names)
        .collect();

    // Step 3: Collect already-imported symbols
    let existing_imports = extract_existing_import_names(&resolved);

    // Step 4: Find unresolved = used - local - imported - builtins
    let unresolved: Vec<String> = used_types
        .into_iter()
        .filter(|name| !local_symbols.contains(name) && !existing_imports.contains(name))
        .filter(|name| !is_builtin(name, &ext))
        .collect();

    // Step 5: Batch lookup via SymbolIndex (SQLite) — much faster than per-name find_symbol
    let insert_line = find_import_insert_line(&source, &ext);
    let mut suggestions = Vec::new();
    let index = SymbolIndex::new(project.clone());

    for name in &unresolved {
        if let Ok(matches) = index.find_symbol(name, None, false, true, 3) {
            // Skip if only found in the same file
            let external: Vec<_> = matches
                .iter()
                .filter(|m| m.file_path != file_path)
                .collect();
            let best_ref = external.first().copied().or(matches.first());
            if let Some(best) = best_ref {
                let import_stmt = generate_import_statement(name, &best.file_path, &ext);
                suggestions.push(ImportSuggestion {
                    symbol_name: name.clone(),
                    source_file: best.file_path.clone(),
                    import_statement: import_stmt,
                    insert_line,
                    confidence: if external.len() == 1 { 0.95 } else { 0.7 },
                });
            }
        }
    }

    let result = MissingImportAnalysis {
        file_path: file_path.to_string(),
        unresolved_symbols: unresolved,
        suggestions,
    };

    // Store in cache, evict oldest if at capacity
    if let Ok(mut cache) = IMPORT_ANALYSIS_CACHE.lock() {
        if cache.len() >= IMPORT_CACHE_CAPACITY
            && let Some(&oldest_key) = cache.keys().next()
        {
            cache.remove(&oldest_key);
        }
        cache.insert(cache_key, result.clone());
    }

    Ok(result)
}

/// Add an import statement to a file at the correct position.
pub fn add_import(
    project: &ProjectRoot,
    file_path: &str,
    import_statement: &str,
) -> Result<(String, crate::edit_transaction::ApplyEvidence)> {
    let resolved = project.resolve(file_path)?;
    let content = fs::read_to_string(&resolved)?;
    let ext = resolved
        .extension()
        .and_then(|e| e.to_str())
        .unwrap_or("")
        .to_ascii_lowercase();

    // Check if already imported — synthesize a no-op Applied evidence
    if content.contains(import_statement.trim()) {
        let evidence = crate::edit_transaction::ApplyEvidence {
            status: crate::edit_transaction::ApplyStatus::Applied,
            file_hashes_before: Default::default(),
            file_hashes_after: Default::default(),
            rollback_report: vec![],
            modified_files: 0,
            edit_count: 0,
        };
        return Ok((content, evidence));
    }

    let insert_line = find_import_insert_line(&content, &ext);
    let mut lines: Vec<&str> = content.lines().collect();
    let insert_idx = (insert_line - 1).min(lines.len());
    lines.insert(insert_idx, import_statement.trim());

    let mut new_content = lines.join("\n");
    if content.ends_with('\n') {
        new_content.push('\n');
    }

    let relative_path = file_path;
    let evidence = match crate::edit_transaction::apply_full_write_with_evidence(
        project,
        relative_path,
        &new_content,
    ) {
        Ok(ev) => ev,
        Err(crate::edit_transaction::ApplyError::ApplyFailed {
            source: _,
            evidence,
        }) => evidence,
        Err(other) => return Err(anyhow::Error::msg(other.to_string())),
    };
    Ok((new_content, evidence))
}

// ── Helpers ──────────────────────────────────────────────────────────────

/// Collect type candidates from AST — only real type identifiers, not strings/comments.
fn collect_type_candidates_ast(file_path: &Path, source: &str) -> Result<Vec<String>> {
    let Some(config) = language_for_path(file_path) else {
        // Fallback to regex for unsupported languages
        return Ok(collect_type_candidates_regex(source));
    };

    let mut parser = Parser::new();
    parser.set_language(&config.language)?;
    let Some(tree) = parser.parse(source, None) else {
        return Ok(collect_type_candidates_regex(source));
    };

    let source_bytes = source.as_bytes();
    let mut seen = HashSet::new();
    let mut result = Vec::new();
    collect_type_nodes(tree.root_node(), source_bytes, &mut seen, &mut result);
    Ok(result)
}

fn collect_type_nodes(
    node: Node,
    source: &[u8],
    seen: &mut HashSet<String>,
    out: &mut Vec<String>,
) {
    let kind = node.kind();

    // Skip comments and strings entirely
    if matches!(
        kind,
        "comment"
            | "line_comment"
            | "block_comment"
            | "string"
            | "string_literal"
            | "template_string"
            | "raw_string_literal"
            | "interpreted_string_literal"
    ) {
        return;
    }

    // Collect type identifiers and uppercase identifiers in type positions
    if kind == "type_identifier" || kind == "identifier" {
        let text = std::str::from_utf8(&source[node.byte_range()]).unwrap_or("");
        if !text.is_empty()
            && text
                .chars()
                .next()
                .map(|c| c.is_uppercase())
                .unwrap_or(false)
            && !is_keyword(text)
            && seen.insert(text.to_string())
        {
            out.push(text.to_string());
        }
    }

    for i in 0..node.child_count() {
        if let Some(child) = node.child(i) {
            collect_type_nodes(child, source, seen, out);
        }
    }
}

/// Regex fallback for unsupported languages.
fn collect_type_candidates_regex(source: &str) -> Vec<String> {
    let re = &*TYPE_CANDIDATE_RE;
    let mut seen = HashSet::new();
    let mut result = Vec::new();
    for line in source.lines() {
        let trimmed = line.trim();
        if trimmed.starts_with('#') || trimmed.starts_with("//") || trimmed.starts_with("/*") {
            continue;
        }
        for cap in re.find_iter(line) {
            let name = cap.as_str().to_string();
            if !is_keyword(&name) && seen.insert(name.clone()) {
                result.push(name);
            }
        }
    }
    result
}

/// Extract names that are already imported.
fn extract_existing_import_names(path: &Path) -> HashSet<String> {
    let raw_imports = extract_imports_for_file(path);
    let mut names = HashSet::new();
    for imp in &raw_imports {
        // Extract last segment: "from foo import Bar" → "Bar", "import foo.Bar" → "Bar"
        if let Some(last) = imp.rsplit('.').next() {
            names.insert(last.to_string());
        }
        // Also try extracting from "from X import Y" patterns
        if let Some(pos) = imp.find(" import ") {
            let after = &imp[pos + 8..];
            for part in after.split(',') {
                let name = part.trim().split(" as ").next().unwrap_or("").trim();
                if !name.is_empty() {
                    names.insert(name.to_string());
                }
            }
        }
    }
    names
}

/// Find the line number where new imports should be inserted.
fn find_import_insert_line(source: &str, ext: &str) -> usize {
    let mut last_import_line = 0;
    let mut in_docstring = false;

    for (i, line) in source.lines().enumerate() {
        let trimmed = line.trim();

        // Skip Python docstrings
        if trimmed.contains("\"\"\"") || trimmed.contains("'''") {
            in_docstring = !in_docstring;
            continue;
        }
        if in_docstring {
            continue;
        }

        let is_import = match ext {
            "py" => trimmed.starts_with("import ") || trimmed.starts_with("from "),
            "ts" | "tsx" | "js" | "jsx" | "mjs" | "cjs" => {
                trimmed.starts_with("import ") || trimmed.starts_with("import{")
            }
            "java" | "kt" | "kts" => trimmed.starts_with("import "),
            "go" => trimmed.starts_with("import ") || trimmed == "import (",
            "rs" => trimmed.starts_with("use "),
            _ => false,
        };

        if is_import {
            last_import_line = i + 1;
        }
    }

    // If no imports found, insert after package/module declaration or at top
    if last_import_line == 0 {
        for (i, line) in source.lines().enumerate() {
            let trimmed = line.trim();
            if trimmed.starts_with("package ")
                || trimmed.starts_with("module ")
                || (trimmed.starts_with('#') && trimmed.contains("!"))
            {
                return i + 2; // After package + blank line
            }
        }
        return 1;
    }

    last_import_line + 1
}

/// Generate a language-appropriate import statement.
fn generate_import_statement(symbol_name: &str, source_file: &str, target_ext: &str) -> String {
    let module = source_file
        .trim_end_matches(".py")
        .trim_end_matches(".ts")
        .trim_end_matches(".tsx")
        .trim_end_matches(".js")
        .trim_end_matches(".jsx")
        .trim_end_matches(".java")
        .trim_end_matches(".kt")
        .trim_end_matches(".rs")
        .trim_end_matches(".go")
        .replace('/', ".");

    match target_ext {
        "py" => format!("from {module} import {symbol_name}"),
        "ts" | "tsx" | "js" | "jsx" | "mjs" | "cjs" => {
            let rel_path = format!(
                "./{}",
                source_file
                    .trim_end_matches(".ts")
                    .trim_end_matches(".tsx")
                    .trim_end_matches(".js")
            );
            format!("import {{ {} }} from '{}';", symbol_name, rel_path)
        }
        "java" => format!("import {};", module),
        "kt" | "kts" => format!("import {}", module),
        "rs" => format!("use crate::{};", module.replace('.', "::")),
        "go" => format!("import \"{}\"", source_file.trim_end_matches(".go")),
        _ => format!("// import {} from {}", symbol_name, source_file),
    }
}

fn flatten_names(symbol: SymbolInfo) -> Vec<String> {
    let mut names = vec![symbol.name.clone()];
    for child in symbol.children {
        names.extend(flatten_names(child));
    }
    names
}

fn is_keyword(name: &str) -> bool {
    matches!(
        name,
        "True"
            | "False"
            | "None"
            | "Self"
            | "String"
            | "Result"
            | "Option"
            | "Vec"
            | "HashMap"
            | "HashSet"
            | "Object"
            | "Array"
            | "Map"
            | "Set"
            | "Promise"
            | "Error"
            | "TypeError"
            | "ValueError"
            | "Exception"
            | "RuntimeError"
            | "Boolean"
            | "Integer"
            | "Float"
            | "Double"
            | "NULL"
            | "EOF"
            | "TODO"
            | "FIXME"
            | "HACK"
    )
}

fn is_builtin(name: &str, ext: &str) -> bool {
    if is_keyword(name) {
        return true;
    }
    match ext {
        "py" => matches!(
            name,
            "int"
                | "str"
                | "float"
                | "bool"
                | "list"
                | "dict"
                | "tuple"
                | "set"
                | "Type"
                | "Optional"
                | "List"
                | "Dict"
                | "Tuple"
                | "Set"
                | "Any"
                | "Union"
                | "Callable"
        ),
        "ts" | "tsx" | "js" | "jsx" => matches!(
            name,
            "Date"
                | "RegExp"
                | "JSON"
                | "Math"
                | "Number"
                | "Console"
                | "Window"
                | "Document"
                | "Element"
                | "HTMLElement"
                | "Event"
                | "Response"
                | "Request"
                | "Partial"
                | "Readonly"
                | "Record"
                | "Pick"
                | "Omit"
        ),
        "java" | "kt" => matches!(
            name,
            "System"
                | "Math"
                | "Thread"
                | "Class"
                | "Comparable"
                | "Iterable"
                | "Iterator"
                | "Override"
                | "Deprecated"
                | "Test"
                | "Suppress"
        ),
        "rs" => matches!(
            name,
            "Ok" | "Err"
                | "Some"
                | "Copy"
                | "Clone"
                | "Debug"
                | "Default"
                | "Display"
                | "From"
                | "Into"
                | "Send"
                | "Sync"
                | "Sized"
                | "Drop"
                | "Fn"
                | "FnMut"
                | "FnOnce"
                | "Box"
                | "Rc"
                | "Arc"
                | "Mutex"
                | "RwLock"
                | "Pin"
                | "Serialize"
                | "Deserialize"
                | "Regex"
                | "Path"
                | "PathBuf"
                | "File"
                | "Read"
                | "Write"
                | "BufRead"
                | "BufReader"
                | "BufWriter"
                | "WalkDir"
                | "Context"
                | "Cow"
                | "PhantomData"
                | "ManuallyDrop"
        ),
        _ => false,
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::ProjectRoot;

    fn make_fixture() -> (std::path::PathBuf, ProjectRoot) {
        let dir = std::env::temp_dir().join(format!(
            "codelens-autoimport-{}",
            std::time::SystemTime::now()
                .duration_since(std::time::UNIX_EPOCH)
                .unwrap()
                .as_nanos()
        ));
        fs::create_dir_all(dir.join("src")).unwrap();
        fs::write(
            dir.join("src/models.py"),
            "class UserModel:\n    def __init__(self, name):\n        self.name = name\n",
        )
        .unwrap();
        fs::write(
            dir.join("src/service.py"),
            "class UserService:\n    def get(self):\n        return UserModel()\n",
        )
        .unwrap();
        let project = ProjectRoot::new(&dir).unwrap();
        (dir, project)
    }

    #[test]
    fn detects_unresolved_type() {
        let (_dir, project) = make_fixture();
        let result = analyze_missing_imports(&project, "src/service.py").unwrap();
        assert!(
            result.unresolved_symbols.contains(&"UserModel".to_string()),
            "should detect UserModel as unresolved: {:?}",
            result.unresolved_symbols
        );
    }

    #[test]
    fn suggests_import_for_unresolved() {
        let (_dir, project) = make_fixture();
        let result = analyze_missing_imports(&project, "src/service.py").unwrap();
        let suggestion = result
            .suggestions
            .iter()
            .find(|s| s.symbol_name == "UserModel");
        assert!(
            suggestion.is_some(),
            "should suggest import for UserModel: {:?}",
            result.suggestions
        );
        let s = suggestion.unwrap();
        assert!(
            s.import_statement.contains("UserModel"),
            "import statement should mention UserModel: {}",
            s.import_statement
        );
        assert!(s.confidence > 0.5);
    }

    #[test]
    fn does_not_suggest_locally_defined() {
        let (_dir, project) = make_fixture();
        let result = analyze_missing_imports(&project, "src/models.py").unwrap();
        assert!(
            !result.unresolved_symbols.contains(&"UserModel".to_string()),
            "locally defined UserModel should not be unresolved"
        );
    }

    #[test]
    fn add_import_inserts_at_correct_position() {
        use crate::edit_transaction::ApplyStatus;

        let dir = std::env::temp_dir().join(format!(
            "codelens-addimport-{}",
            std::time::SystemTime::now()
                .duration_since(std::time::UNIX_EPOCH)
                .unwrap()
                .as_nanos()
        ));
        fs::create_dir_all(&dir).unwrap();
        fs::write(
            dir.join("test.py"),
            "import os\nimport sys\n\ndef main():\n    pass\n",
        )
        .unwrap();
        let project = ProjectRoot::new(&dir).unwrap();
        let (result, evidence) =
            add_import(&project, "test.py", "from models import User").unwrap();
        assert_eq!(
            evidence.status,
            ApplyStatus::Applied,
            "evidence should be Applied"
        );
        let lines: Vec<&str> = result.lines().collect();
        // Should be inserted after existing imports (line 3)
        assert!(
            lines.contains(&"from models import User"),
            "should contain new import: {:?}",
            lines
        );
        let import_idx = lines
            .iter()
            .position(|l| *l == "from models import User")
            .unwrap();
        let sys_idx = lines.iter().position(|l| *l == "import sys").unwrap();
        assert!(
            import_idx > sys_idx,
            "new import should be after existing imports"
        );
    }

    #[test]
    fn skip_already_imported() {
        let dir = std::env::temp_dir().join(format!(
            "codelens-skipimport-{}",
            std::time::SystemTime::now()
                .duration_since(std::time::UNIX_EPOCH)
                .unwrap()
                .as_nanos()
        ));
        fs::create_dir_all(&dir).unwrap();
        fs::write(
            dir.join("test.py"),
            "from models import User\n\nx = User()\n",
        )
        .unwrap();
        let project = ProjectRoot::new(&dir).unwrap();
        let (result, _evidence) =
            add_import(&project, "test.py", "from models import User").unwrap();
        // Should not duplicate
        assert_eq!(
            result.matches("from models import User").count(),
            1,
            "should not duplicate import"
        );
    }

    #[test]
    fn find_import_insert_line_python() {
        let source = "import os\nimport sys\n\ndef main():\n    pass\n";
        assert_eq!(find_import_insert_line(source, "py"), 3);
    }

    #[test]
    fn find_import_insert_line_empty() {
        let source = "def main():\n    pass\n";
        assert_eq!(find_import_insert_line(source, "py"), 1);
    }

    #[test]
    fn generate_python_import() {
        let stmt = generate_import_statement("UserModel", "src/models.py", "py");
        assert_eq!(stmt, "from src.models import UserModel");
    }

    #[test]
    fn generate_typescript_import() {
        let stmt = generate_import_statement("UserService", "src/service.ts", "ts");
        assert!(stmt.contains("import { UserService }"));
        assert!(stmt.contains("'./src/service'"));
    }
}