denoland-deno/cli/tools/coverage.rs

// Copyright 2018-2021 the Deno authors. All rights reserved. MIT license.

use crate::colors;
use crate::emit;
use crate::flags::CoverageFlags;
use crate::flags::Flags;
use crate::fs_util::collect_files;
use crate::proc_state::ProcState;
use crate::source_maps::SourceMapGetter;
use crate::tools::fmt::format_json;

use deno_ast::MediaType;
use deno_ast::ModuleSpecifier;
use deno_core::error::AnyError;
use deno_core::serde_json;
use deno_core::url::Url;
use deno_core::LocalInspectorSession;
use deno_runtime::permissions::Permissions;
use regex::Regex;
use serde::Deserialize;
use serde::Serialize;
use sourcemap::SourceMap;
use std::fs;
use std::fs::File;
use std::io::BufWriter;
use std::io::Write;
use std::path::PathBuf;
use text_lines::TextLines;
use uuid::Uuid;

#[derive(Debug, Serialize, Deserialize, Clone)]
#[serde(rename_all = "camelCase")]
struct CoverageRange {
  /// Start byte index.
  start_offset: usize,
  /// End byte index.
  end_offset: usize,
  count: usize,
}

#[derive(Debug, Serialize, Deserialize, Clone)]
#[serde(rename_all = "camelCase")]
struct FunctionCoverage {
  function_name: String,
  ranges: Vec<CoverageRange>,
  is_block_coverage: bool,
}

#[derive(Debug, Serialize, Deserialize, Clone)]
#[serde(rename_all = "camelCase")]
struct ScriptCoverage {
  script_id: String,
  url: String,
  functions: Vec<FunctionCoverage>,
}

#[derive(Debug, Serialize, Deserialize)]
#[serde(rename_all = "camelCase")]
struct StartPreciseCoverageParameters {
  call_count: bool,
  detailed: bool,
  allow_triggered_updates: bool,
}

#[derive(Debug, Serialize, Deserialize)]
#[serde(rename_all = "camelCase")]
struct StartPreciseCoverageReturnObject {
  timestamp: f64,
}

#[derive(Debug, Serialize, Deserialize)]
#[serde(rename_all = "camelCase")]
struct TakePreciseCoverageReturnObject {
  result: Vec<ScriptCoverage>,
  timestamp: f64,
}

pub struct CoverageCollector {
  pub dir: PathBuf,
  session: LocalInspectorSession,
}

impl CoverageCollector {
  pub fn new(dir: PathBuf, session: LocalInspectorSession) -> Self {
    Self { dir, session }
  }

  async fn enable_debugger(&mut self) -> Result<(), AnyError> {
    self.session.post_message("Debugger.enable", None).await?;
    Ok(())
  }

  async fn enable_profiler(&mut self) -> Result<(), AnyError> {
    self.session.post_message("Profiler.enable", None).await?;
    Ok(())
  }

  async fn disable_debugger(&mut self) -> Result<(), AnyError> {
    self.session.post_message("Debugger.disable", None).await?;
    Ok(())
  }

  async fn disable_profiler(&mut self) -> Result<(), AnyError> {
    self.session.post_message("Profiler.disable", None).await?;
    Ok(())
  }

  async fn start_precise_coverage(
    &mut self,
    parameters: StartPreciseCoverageParameters,
  ) -> Result<StartPreciseCoverageReturnObject, AnyError> {
    let parameters_value = serde_json::to_value(parameters)?;
    let return_value = self
      .session
      .post_message("Profiler.startPreciseCoverage", Some(parameters_value))
      .await?;

    let return_object = serde_json::from_value(return_value)?;

    Ok(return_object)
  }

  async fn take_precise_coverage(
    &mut self,
  ) -> Result<TakePreciseCoverageReturnObject, AnyError> {
    let return_value = self
      .session
      .post_message("Profiler.takePreciseCoverage", None)
      .await?;

    let return_object = serde_json::from_value(return_value)?;

    Ok(return_object)
  }

  pub async fn start_collecting(&mut self) -> Result<(), AnyError> {
    self.enable_debugger().await?;
    self.enable_profiler().await?;
    self
      .start_precise_coverage(StartPreciseCoverageParameters {
        call_count: true,
        detailed: true,
        allow_triggered_updates: false,
      })
      .await?;

    Ok(())
  }

  pub async fn stop_collecting(&mut self) -> Result<(), AnyError> {
    fs::create_dir_all(&self.dir)?;

    let script_coverages = self.take_precise_coverage().await?.result;
    for script_coverage in script_coverages {
      let filename = format!("{}.json", Uuid::new_v4());
      let filepath = self.dir.join(filename);

      let mut out = BufWriter::new(File::create(filepath)?);
      let coverage = serde_json::to_string(&script_coverage)?;
      let formated_coverage =
        format_json(&coverage, &Default::default()).unwrap_or(coverage);

      out.write_all(formated_coverage.as_bytes())?;
      out.flush()?;
    }

    self.disable_debugger().await?;
    self.disable_profiler().await?;

    Ok(())
  }
}

struct BranchCoverageItem {
  line_index: usize,
  block_number: usize,
  branch_number: usize,
  taken: Option<usize>,
  is_hit: bool,
}

struct FunctionCoverageItem {
  name: String,
  line_index: usize,
  execution_count: usize,
}

struct CoverageReport {
  url: ModuleSpecifier,
  named_functions: Vec<FunctionCoverageItem>,
  branches: Vec<BranchCoverageItem>,
  found_lines: Vec<(usize, usize)>,
}

fn generate_coverage_report(
  script_coverage: &ScriptCoverage,
  script_source: &str,
  maybe_source_map: &Option<Vec<u8>>,
) -> CoverageReport {
  let maybe_source_map = maybe_source_map
    .as_ref()
    .map(|source_map| SourceMap::from_slice(source_map).unwrap());
  let text_lines = TextLines::new(script_source);

  let comment_spans = deno_ast::lex(script_source, MediaType::JavaScript)
    .into_iter()
    .filter(|item| {
      matches!(item.inner, deno_ast::TokenOrComment::Comment { .. })
    })
    .map(|item| item.span)
    .collect::<Vec<_>>();

  let url = Url::parse(&script_coverage.url).unwrap();
  let mut coverage_report = CoverageReport {
    url,
    named_functions: Vec::with_capacity(
      script_coverage
        .functions
        .iter()
        .filter(|f| !f.function_name.is_empty())
        .count(),
    ),
    branches: Vec::new(),
    found_lines: Vec::new(),
  };

  for function in &script_coverage.functions {
    if function.function_name.is_empty() {
      continue;
    }

    let source_line_index =
      text_lines.line_index(function.ranges[0].start_offset);
    let line_index = if let Some(source_map) = maybe_source_map.as_ref() {
      source_map
        .tokens()
        .find(|token| token.get_dst_line() as usize == source_line_index)
        .map(|token| token.get_src_line() as usize)
        .unwrap_or(0)
    } else {
      source_line_index
    };

    coverage_report.named_functions.push(FunctionCoverageItem {
      name: function.function_name.clone(),
      line_index,
      execution_count: function.ranges[0].count,
    });
  }

  for (block_number, function) in script_coverage.functions.iter().enumerate() {
    let block_hits = function.ranges[0].count;
    for (branch_number, range) in function.ranges[1..].iter().enumerate() {
      let source_line_index = text_lines.line_index(range.start_offset);
      let line_index = if let Some(source_map) = maybe_source_map.as_ref() {
        source_map
          .tokens()
          .find(|token| token.get_dst_line() as usize == source_line_index)
          .map(|token| token.get_src_line() as usize)
          .unwrap_or(0)
      } else {
        source_line_index
      };

      // From https://manpages.debian.org/unstable/lcov/geninfo.1.en.html:
      //
      // Block number and branch number are gcc internal IDs for the branch. Taken is either '-'
      // if the basic block containing the branch was never executed or a number indicating how
      // often that branch was taken.
      //
      // However with the data we get from v8 coverage profiles it seems we can't actually hit
      // this as appears it won't consider any nested branches it hasn't seen but its here for
      // the sake of accuracy.
      let taken = if block_hits > 0 {
        Some(range.count)
      } else {
        None
      };

      coverage_report.branches.push(BranchCoverageItem {
        line_index,
        block_number,
        branch_number,
        taken,
        is_hit: range.count > 0,
      })
    }
  }

  // TODO(caspervonb): collect uncovered ranges on the lines so that we can highlight specific
  // parts of a line in color (word diff style) instead of the entire line.
  let mut line_counts = Vec::with_capacity(text_lines.lines_count());
  for line_index in 0..text_lines.lines_count() {
    let line_start_offset = text_lines.line_start(line_index);
    let line_end_offset = text_lines.line_end(line_index);
    let ignore = comment_spans.iter().any(|span| {
      (span.lo.0 as usize) <= line_start_offset
        && (span.hi.0 as usize) >= line_end_offset
    }) || script_source[line_start_offset..line_end_offset]
      .trim()
      .is_empty();
    let mut count = 0;

    if ignore {
      count = 1;
    } else {
      // Count the hits of ranges that include the entire line which will always be at-least one
      // as long as the code has been evaluated.
      for function in &script_coverage.functions {
        for range in &function.ranges {
          if range.start_offset <= line_start_offset
            && range.end_offset >= line_end_offset
          {
            count += range.count;
          }
        }
      }

      // We reset the count if any block with a zero count overlaps with the line range.
      for function in &script_coverage.functions {
        for range in &function.ranges {
          if range.count > 0 {
            continue;
          }

          let overlaps = range.start_offset < line_end_offset
            && range.end_offset > line_start_offset;
          if overlaps {
            count = 0;
          }
        }
      }
    }

    line_counts.push(count);
  }

  coverage_report.found_lines =
    if let Some(source_map) = maybe_source_map.as_ref() {
      let mut found_lines = line_counts
        .iter()
        .enumerate()
        .map(|(index, count)| {
          // get all the mappings from this destination line to a different src line
          let mut results = source_map
            .tokens()
            .filter(move |token| token.get_dst_line() as usize == index)
            .map(move |token| (token.get_src_line() as usize, *count))
            .collect::<Vec<_>>();
          // only keep the results that point at different src lines
          results.sort_unstable_by_key(|(index, _)| *index);
          results.dedup_by_key(|(index, _)| *index);
          results.into_iter()
        })
        .flatten()
        .collect::<Vec<(usize, usize)>>();

      found_lines.sort_unstable_by_key(|(index, _)| *index);
      // combine duplicated lines
      for i in (1..found_lines.len()).rev() {
        if found_lines[i].0 == found_lines[i - 1].0 {
          found_lines[i - 1].1 += found_lines[i].1;
          found_lines.remove(i);
        }
      }
      found_lines
    } else {
      line_counts
        .into_iter()
        .enumerate()
        .map(|(index, count)| (index, count))
        .collect::<Vec<(usize, usize)>>()
    };

  coverage_report
}

enum CoverageReporterKind {
  Pretty,
  Lcov,
}

fn create_reporter(
  kind: CoverageReporterKind,
) -> Box<dyn CoverageReporter + Send> {
  match kind {
    CoverageReporterKind::Lcov => Box::new(LcovCoverageReporter::new()),
    CoverageReporterKind::Pretty => Box::new(PrettyCoverageReporter::new()),
  }
}

trait CoverageReporter {
  fn report(&mut self, coverage_report: &CoverageReport, file_text: &str);

  fn done(&mut self);
}

struct LcovCoverageReporter {}

impl LcovCoverageReporter {
  pub fn new() -> LcovCoverageReporter {
    LcovCoverageReporter {}
  }
}

impl CoverageReporter for LcovCoverageReporter {
  fn report(&mut self, coverage_report: &CoverageReport, _file_text: &str) {
    let file_path = coverage_report
      .url
      .to_file_path()
      .ok()
      .map(|p| p.to_str().map(|p| p.to_string()))
      .flatten()
      .unwrap_or_else(|| coverage_report.url.to_string());
    println!("SF:{}", file_path);

    for function in &coverage_report.named_functions {
      println!("FN:{},{}", function.line_index + 1, function.name);
    }

    for function in &coverage_report.named_functions {
      println!("FNDA:{},{}", function.execution_count, function.name);
    }

    let functions_found = coverage_report.named_functions.len();
    println!("FNF:{}", functions_found);
    let functions_hit = coverage_report
      .named_functions
      .iter()
      .filter(|f| f.execution_count > 0)
      .count();
    println!("FNH:{}", functions_hit);

    for branch in &coverage_report.branches {
      let taken = if let Some(taken) = &branch.taken {
        taken.to_string()
      } else {
        "-".to_string()
      };

      println!(
        "BRDA:{},{},{},{}",
        branch.line_index + 1,
        branch.block_number,
        branch.branch_number,
        taken
      );
    }

    let branches_found = coverage_report.branches.len();
    println!("BRF:{}", branches_found);
    let branches_hit =
      coverage_report.branches.iter().filter(|b| b.is_hit).count();
    println!("BRH:{}", branches_hit);

    for (index, count) in &coverage_report.found_lines {
      println!("DA:{},{}", index + 1, count);
    }

    let lines_hit = coverage_report
      .found_lines
      .iter()
      .filter(|(_, count)| *count != 0)
      .count();
    println!("LH:{}", lines_hit);

    let lines_found = coverage_report.found_lines.len();
    println!("LF:{}", lines_found);

    println!("end_of_record");
  }

  fn done(&mut self) {}
}

struct PrettyCoverageReporter {}

impl PrettyCoverageReporter {
  pub fn new() -> PrettyCoverageReporter {
    PrettyCoverageReporter {}
  }
}

impl CoverageReporter for PrettyCoverageReporter {
  fn report(&mut self, coverage_report: &CoverageReport, file_text: &str) {
    let lines = file_text.split('\n').collect::<Vec<_>>();
    print!("cover {} ... ", coverage_report.url);

    let hit_lines = coverage_report
      .found_lines
      .iter()
      .filter(|(_, count)| *count > 0)
      .map(|(index, _)| *index);

    let missed_lines = coverage_report
      .found_lines
      .iter()
      .filter(|(_, count)| *count == 0)
      .map(|(index, _)| *index);

    let lines_found = coverage_report.found_lines.len();
    let lines_hit = hit_lines.count();
    let line_ratio = lines_hit as f32 / lines_found as f32;

    let line_coverage =
      format!("{:.3}% ({}/{})", line_ratio * 100.0, lines_hit, lines_found);

    if line_ratio >= 0.9 {
      println!("{}", colors::green(&line_coverage));
    } else if line_ratio >= 0.75 {
      println!("{}", colors::yellow(&line_coverage));
    } else {
      println!("{}", colors::red(&line_coverage));
    }

    let mut last_line = None;
    for line_index in missed_lines {
      const WIDTH: usize = 4;
      const SEPERATOR: &str = "|";

      // Put a horizontal separator between disjoint runs of lines
      if let Some(last_line) = last_line {
        if last_line + 1 != line_index {
          let dash = colors::gray("-".repeat(WIDTH + 1));
          println!("{}{}{}", dash, colors::gray(SEPERATOR), dash);
        }
      }

      println!(
        "{:width$} {} {}",
        line_index + 1,
        colors::gray(SEPERATOR),
        colors::red(&lines[line_index]),
        width = WIDTH
      );

      last_line = Some(line_index);
    }
  }

  fn done(&mut self) {}
}

fn collect_coverages(
  files: Vec<PathBuf>,
  ignore: Vec<PathBuf>,
) -> Result<Vec<ScriptCoverage>, AnyError> {
  let mut coverages: Vec<ScriptCoverage> = Vec::new();
  let file_paths = collect_files(&files, &ignore, |file_path| {
    file_path.extension().map_or(false, |ext| ext == "json")
  })?;

  for file_path in file_paths {
    let json = fs::read_to_string(file_path.as_path())?;
    let new_coverage: ScriptCoverage = serde_json::from_str(&json)?;

    let existing_coverage =
      coverages.iter_mut().find(|x| x.url == new_coverage.url);

    if let Some(existing_coverage) = existing_coverage {
      for new_function in new_coverage.functions {
        let existing_function = existing_coverage
          .functions
          .iter_mut()
          .find(|x| x.function_name == new_function.function_name);

        if let Some(existing_function) = existing_function {
          for new_range in new_function.ranges {
            let existing_range =
              existing_function.ranges.iter_mut().find(|x| {
                x.start_offset == new_range.start_offset
                  && x.end_offset == new_range.end_offset
              });

            if let Some(existing_range) = existing_range {
              existing_range.count += new_range.count;
            } else {
              existing_function.ranges.push(new_range);
            }
          }
        } else {
          existing_coverage.functions.push(new_function);
        }
      }
    } else {
      coverages.push(new_coverage);
    }
  }

  coverages.sort_by_key(|k| k.url.clone());

  Ok(coverages)
}

fn filter_coverages(
  coverages: Vec<ScriptCoverage>,
  include: Vec<String>,
  exclude: Vec<String>,
) -> Vec<ScriptCoverage> {
  let include: Vec<Regex> =
    include.iter().map(|e| Regex::new(e).unwrap()).collect();

  let exclude: Vec<Regex> =
    exclude.iter().map(|e| Regex::new(e).unwrap()).collect();

  coverages
    .into_iter()
    .filter(|e| {
      let is_internal = e.url.starts_with("deno:")
        || e.url.ends_with("__anonymous__")
        || e.url.ends_with("$deno$test.js");

      let is_included = include.iter().any(|p| p.is_match(&e.url));
      let is_excluded = exclude.iter().any(|p| p.is_match(&e.url));

      (include.is_empty() || is_included) && !is_excluded && !is_internal
    })
    .collect::<Vec<ScriptCoverage>>()
}

pub async fn cover_files(
  flags: Flags,
  coverage_flags: CoverageFlags,
) -> Result<(), AnyError> {
  let ps = ProcState::build(flags).await?;

  let script_coverages =
    collect_coverages(coverage_flags.files, coverage_flags.ignore)?;
  let script_coverages = filter_coverages(
    script_coverages,
    coverage_flags.include,
    coverage_flags.exclude,
  );

  let reporter_kind = if coverage_flags.lcov {
    CoverageReporterKind::Lcov
  } else {
    CoverageReporterKind::Pretty
  };

  let mut reporter = create_reporter(reporter_kind);

  for script_coverage in script_coverages {
    let module_specifier =
      deno_core::resolve_url_or_path(&script_coverage.url)?;
    ps.prepare_module_load(
      vec![module_specifier.clone()],
      false,
      emit::TypeLib::UnstableDenoWindow,
      Permissions::allow_all(),
      Permissions::allow_all(),
      false,
    )
    .await?;

    let module_source = ps.load(module_specifier.clone(), None, false)?;
    let script_source = &module_source.code;

    let maybe_source_map = ps.get_source_map(&script_coverage.url);
    let maybe_cached_source = ps
      .file_fetcher
      .get_source(&module_specifier)
      .map(|f| f.source);

    let coverage_report = generate_coverage_report(
      &script_coverage,
      script_source,
      &maybe_source_map,
    );

    let file_text = if let Some(original_source) =
      maybe_source_map.and(maybe_cached_source.as_ref())
    {
      original_source.as_str()
    } else {
      script_source
    };

    reporter.report(&coverage_report, file_text);
  }

  reporter.done();

  Ok(())
}