feat: add processinfo index, add externalId (#1055)

If a NYC_PROCESSINFO_EXTERNAL_ID environment variable is set, then it is
saved in the processinfo as `externalId`.

Furthermore, when this file is generated, some additional helpful
metadata is memoized to the processinfo json files, to minimize the cost
of repeated generation.  (This isn't necessarily a breaking change, but
it is an update to the de facto schema for those files.)

As soon as possible, index generation and process tree display should be
migrated out to a new 'istanbul-lib-processinfo' library.

This opens the door to add features in the v14 release family to improve
support for partial/resumed test runs and file watching.

- When a process is run with --clean=false and a previously seen
  externalId, clear away all the coverage files in the set for that
  externalId.
- When a file is changed, a test runner can use the index to determine
  which tests (by externalId) ought to be re-run.
- Adds a NYC_PROCESS_ID to environment
- Adds `parent` to processInfo object, a uuid referring to parent.
- Rebase onto processinfo-numeric-pids branch
- Avoid re-writing the processinfo/{uuid}.json files
- Update process tree output to rely on process index instead of
  duplicating effort.

BREAKING CHANGE: This adds a file named 'index.json' to the
.nyc_output/processinfo directory, which has a different format from the
other files in this dir.

Author: isaacs

bin/nyc.js

@@ -65,6 +65,10 @@ if ([
   ), function (done) {
     var mainChildExitCode = process.exitCode
 
+    if (argv.showProcessTree || argv.buildProcessTree) {
+      nyc.writeProcessIndex()
+    }
+
     if (argv.checkCoverage) {
       nyc.checkCoverage({
         lines: argv.lines,

bin/wrap.js

@@ -8,8 +8,13 @@ config.isChildProcess = true
 config._processInfo = {
   pid: process.pid,
   ppid: process.ppid,
+  parent: process.env.NYC_PROCESS_ID || null,
   root: process.env.NYC_ROOT_ID
 }
+if (process.env.NYC_PROCESSINFO_EXTERNAL_ID) {
+  config._processInfo.externalId = process.env.NYC_PROCESSINFO_EXTERNAL_ID
+  delete process.env.NYC_PROCESSINFO_EXTERNAL_ID
+}
 
 ;(new NYC(config)).wrap()
 

index.js

@@ -312,6 +312,7 @@ NYC.prototype._wrapExit = function () {
 }
 
 NYC.prototype.wrap = function (bin) {
+  process.env.NYC_PROCESS_ID = this.processInfo.uuid
   this._addRequireHooks()
   this._wrapExit()
   this._loadAdditionalModules()
@@ -341,7 +342,7 @@ NYC.prototype.writeCoverageFile = function () {
     coverage = this.sourceMaps.remapCoverage(coverage)
   }
 
-  var id = this.generateUniqueID()
+  var id = this.processInfo.uuid
   var coverageFilename = path.resolve(this.tempDirectory(), id + '.json')
 
   fs.writeFileSync(
@@ -355,6 +356,7 @@ NYC.prototype.writeCoverageFile = function () {
   }
 
   this.processInfo.coverageFilename = coverageFilename
+  this.processInfo.files = Object.keys(coverage)
 
   fs.writeFileSync(
     path.resolve(this.processInfoDirectory(), id + '.json'),
@@ -412,6 +414,80 @@ NYC.prototype.report = function () {
   }
 }
 
+// XXX(@isaacs) Index generation should move to istanbul-lib-processinfo
+NYC.prototype.writeProcessIndex = function () {
+  const dir = this.processInfoDirectory()
+  const pidToUid = new Map()
+  const infoByUid = new Map()
+  const eidToUid = new Map()
+  const infos = fs.readdirSync(dir).filter(f => f !== 'index.json').map(f => {
+    try {
+      const info = JSON.parse(fs.readFileSync(path.resolve(dir, f), 'utf-8'))
+      info.children = []
+      pidToUid.set(info.uuid, info.pid)
+      pidToUid.set(info.pid, info.uuid)
+      infoByUid.set(info.uuid, info)
+      if (info.externalId) {
+        eidToUid.set(info.externalId, info.uuid)
+      }
+      return info
+    } catch (er) {
+      return null
+    }
+  }).filter(Boolean)
+
+  // create all the parent-child links and write back the updated info
+  infos.forEach(info => {
+    if (info.parent) {
+      const parentInfo = infoByUid.get(info.parent)
+      if (parentInfo.children.indexOf(info.uuid) === -1) {
+        parentInfo.children.push(info.uuid)
+      }
+    }
+  })
+
+  // figure out which files were touched by each process.
+  const files = infos.reduce((files, info) => {
+    info.files.forEach(f => {
+      files[f] = files[f] || []
+      files[f].push(info.uuid)
+    })
+    return files
+  }, {})
+
+  // build the actual index!
+  const index = infos.reduce((index, info) => {
+    index.processes[info.uuid] = {}
+    index.processes[info.uuid].parent = info.parent
+    if (info.externalId) {
+      if (index.externalIds[info.externalId]) {
+        throw new Error(`External ID ${info.externalId} used by multiple processes`)
+      }
+      index.processes[info.uuid].externalId = info.externalId
+      index.externalIds[info.externalId] = {
+        root: info.uuid,
+        children: info.children
+      }
+    }
+    index.processes[info.uuid].children = Array.from(info.children)
+    return index
+  }, { processes: {}, files: files, externalIds: {} })
+
+  // flatten the descendant sets of all the externalId procs
+  Object.keys(index.externalIds).forEach(eid => {
+    const { children } = index.externalIds[eid]
+    // push the next generation onto the list so we accumulate them all
+    for (let i = 0; i < children.length; i++) {
+      const nextGen = index.processes[children[i]].children
+      if (nextGen && nextGen.length) {
+        children.push(...nextGen.filter(uuid => children.indexOf(uuid) === -1))
+      }
+    }
+  })
+
+  fs.writeFileSync(path.resolve(dir, 'index.json'), JSON.stringify(index))
+}
+
 NYC.prototype.showProcessTree = function () {
   var processTree = ProcessInfo.buildProcessTree(this._loadProcessInfos())
 
@@ -448,19 +524,25 @@ NYC.prototype._checkCoverage = function (summary, thresholds, file) {
 }
 
 NYC.prototype._loadProcessInfos = function () {
-  var _this = this
-  var files = fs.readdirSync(this.processInfoDirectory())
-
-  return files.map(function (f) {
+  return fs.readdirSync(this.processInfoDirectory()).map(f => {
+    let data
     try {
-      return new ProcessInfo(JSON.parse(fs.readFileSync(
-        path.resolve(_this.processInfoDirectory(), f),
+      data = JSON.parse(fs.readFileSync(
+        path.resolve(this.processInfoDirectory(), f),
         'utf-8'
-      )))
+      ))
     } catch (e) { // handle corrupt JSON output.
-      return {}
+      return null
     }
-  })
+    if (f !== 'index.json') {
+      data.nodes = []
+      data = new ProcessInfo(data)
+    }
+    return { file: path.basename(f, '.json'), data: data }
+  }).filter(Boolean).reduce((infos, info) => {
+    infos[info.file] = info.data
+    return infos
+  }, {})
 }
 
 NYC.prototype.eachReport = function (filenames, iterator, baseDirectory) {

lib/process.js

@@ -1,9 +1,12 @@
 const archy = require('archy')
 const libCoverage = require('istanbul-lib-coverage')
+const uuid = require('uuid/v4')
 
 function ProcessInfo (defaults) {
   defaults = defaults || {}
 
+  this.uuid = null
+  this.parent = null
   this.pid = String(process.pid)
   this.argv = process.argv
   this.execArgv = process.execArgv
@@ -12,13 +15,14 @@ function ProcessInfo (defaults) {
   this.ppid = null
   this.root = null
   this.coverageFilename = null
-  this.nodes = [] // list of children, filled by buildProcessTree()
-
-  this._coverageMap = null
 
   for (var key in defaults) {
     this[key] = defaults[key]
   }
+
+  if (!this.uuid) {
+    this.uuid = uuid()
+  }
 }
 
 Object.defineProperty(ProcessInfo.prototype, 'label', {
@@ -36,29 +40,19 @@ Object.defineProperty(ProcessInfo.prototype, 'label', {
 })
 
 ProcessInfo.buildProcessTree = function (infos) {
-  var treeRoot = new ProcessInfo({ _label: 'nyc' })
-  var nodes = { }
-
-  infos = infos.sort(function (a, b) {
-    return a.time - b.time
-  })
-
-  infos.forEach(function (p) {
-    nodes[p.root + ':' + p.pid] = p
-  })
-
-  infos.forEach(function (p) {
-    if (!p.ppid) {
-      return
+  const treeRoot = new ProcessInfo({ _label: 'nyc', nodes: [] })
+  const index = infos.index
+  for (const id in index.processes) {
+    const node = infos[id]
+    if (!node) {
+      throw new Error(`Invalid entry in processinfo index: ${id}`)
     }
-
-    var parent = nodes[p.root + ':' + p.ppid]
-    if (!parent) {
-      parent = treeRoot
+    const idx = index.processes[id]
+    node.nodes = idx.children.map(id => infos[id]).sort((a, b) => a.time - b.time)
+    if (!node.parent) {
+      treeRoot.nodes.push(node)
     }
-
-    parent.nodes.push(p)
-  })
+  }
 
   return treeRoot
 }

test/processinfo.js

@@ -0,0 +1,95 @@
+const { resolve } = require('path')
+const bin = resolve(__dirname, '../self-coverage/bin/nyc')
+const { spawn } = require('child_process')
+const t = require('tap')
+const rimraf = require('rimraf')
+const node = process.execPath
+const fixturesCLI = resolve(__dirname, './fixtures/cli')
+const tmp = 'processinfo-test'
+const fs = require('fs')
+const resolvedJS = resolve(fixturesCLI, 'selfspawn-fibonacci.js')
+
+rimraf.sync(resolve(fixturesCLI, tmp))
+t.teardown(() => rimraf.sync(resolve(fixturesCLI, tmp)))
+
+t.test('build some processinfo', t => {
+  var args = [
+    bin, '-t', tmp, '--build-process-tree',
+    node, 'selfspawn-fibonacci.js', '5'
+  ]
+  var proc = spawn(process.execPath, args, {
+    cwd: fixturesCLI,
+    env: {
+      PATH: process.env.PATH,
+      NYC_PROCESSINFO_EXTERNAL_ID: 'blorp'
+    }
+  })
+  // don't actually care about the output for this test, just the data
+  proc.stderr.resume()
+  proc.stdout.resume()
+  proc.on('close', (code, signal) => {
+    t.equal(code, 0)
+    t.equal(signal, null)
+    t.end()
+  })
+})
+
+t.test('validate the created processinfo data', t => {
+  const covs = fs.readdirSync(resolve(fixturesCLI, tmp))
+    .filter(f => f !== 'processinfo')
+  t.plan(covs.length * 2)
+
+  covs.forEach(f => {
+    fs.readFile(resolve(fixturesCLI, tmp, f), 'utf8', (er, covjson) => {
+      if (er) {
+        throw er
+      }
+      const covdata = JSON.parse(covjson)
+      t.same(Object.keys(covdata), [resolvedJS])
+      // should have matching processinfo for each cov json
+      const procInfoFile = resolve(fixturesCLI, tmp, 'processinfo', f)
+      fs.readFile(procInfoFile, 'utf8', (er, procInfoJson) => {
+        if (er) {
+          throw er
+        }
+        const procInfoData = JSON.parse(procInfoJson)
+        t.match(procInfoData, {
+          pid: Number,
+          ppid: Number,
+          uuid: f.replace(/\.json$/, ''),
+          argv: [
+            node,
+            resolvedJS,
+            /[1-5]/
+          ],
+          execArgv: [],
+          cwd: fixturesCLI,
+          time: Number,
+          root: /^[0-9a-f]{8}-(?:[0-9a-f]{4}-){3}[0-9a-f]{12}$/,
+          coverageFilename: resolve(fixturesCLI, tmp, f),
+          files: [ resolvedJS ]
+        })
+      })
+    })
+  })
+})
+
+t.test('check out the index', t => {
+  const indexFile = resolve(fixturesCLI, tmp, 'processinfo', 'index.json')
+  const indexJson = fs.readFileSync(indexFile, 'utf-8')
+  const index = JSON.parse(indexJson)
+  const u = /^[0-9a-f]{8}-(?:[0-9a-f]{4}-){3}[0-9a-f]{12}$/
+  t.match(index, {
+    processes: {},
+    files: {
+      [resolvedJS]: [ u, u, u, u, u, u, u, u, u ]
+    },
+    externalIds: {
+      blorp: {
+        root: u,
+        children: [ u, u, u, u, u, u, u, u ]
+      }
+    }
+  })
+  t.end()
+})