Home

Awesome

thetool

<!-- [START badges] -->

Build Status NPM thetool package

<!-- [END badges] -->

thetool is a CLI tool to capture different cpu, memory and other profiles for your node app in Chrome DevTools friendly format.

Quick start

npx thetool -o . -t memorysampling npm run test
# .. open DevTools frontend and do three clicks to get:
<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388239-f9d25580-8024-11e9-8d35-a7ce4f6a4137.png">

Getting Started

thetool works only with Node >= 10.x.

thetool interface is simple as 1-2-3.

  1. Specify output folder using -o flag, e.g. -o . to put output in current folder.
  2. Specify tool using -t, available tools: cpu, memorysampling, memoryallocation, coverage, type, heapsnapshot.
  3. Specify any command to start node, e.g. node index.js or npx thetool or npm run test.

When report is ready, thetool will dump thetool> Report captured in ... message in terminal with a hint how to analyze it.

Why not to use Chrome DevTools directly?

Tool selector

ProblemToolInsightDevTools tab
my app is slowcpuwhere in code does app spend most time?Performance
my app requires too much memorymemorysamplingwhere in code does app allocate most memory?Memory
my app requires too much memorymemoryallocationmost precise version of memorysampling with much bigger overheadMemory
my app requires too much memoryheapsnapshotwhat is inside the heap right now?Memory
my app package is too bigcoveragewhat code was executed and how many times?
my app needs type annotationstypewhat are the types of function arguments and returns?

On-demand tooling

You can use --ondemand flag to profile only part of your app:

  1. Add --ondemand flag to the list of thetool arguments.
  2. Call startTheTool/stopTheTool from your Node scripts (thetool will add these methods to Node context).

startTheTool/stopTheTool methods are asynchronous, so you should await them or chain them using promise.then

Couple examples:

async function main() {
  await startTheTool();
  // code of your app
  await stopTheTool();
}
// .. or using promises..
function main() {
  startTheTool().then(() => {
    // code of your app
  }).then(() => stopTheTool());
}

CPU: Profiler

thetool -o . -t cpu npm run test

To analyze: open Chrome DevTools, to to Performance tab, click load button, select file with data.

<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388318-f55a6c80-8025-11e9-9826-6432b588ed84.png">

Memory: Sampling Profiler

thetool -o . -t memorysampling npm run test

To analyze: open Chrome DevTools, go to Memory tab, click load button, select file with data.

--samplingInterval option is available: average sample interval in bytes, poisson distribution is used for the intervals. The default value is 32768 bytes

<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388239-f9d25580-8024-11e9-8d35-a7ce4f6a4137.png">

Memory: Allocation Profiler

thetool -o . -t memoryallocation npm run test

To analyze: open Chrome DevTools, go to Memory tab, click load button, select file with data.

<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388341-4cf8d800-8026-11e9-9ac5-429a33523ea0.png">

Memory: Heap Snapshot

thetool -o . -t heapsnapshot node -e "captureTheTool.then(captureTheTool).then(captureTheTool)"

Given command will capture three heap snapshots. To analyze: open Chrome DevTools, go to Memory tab, click load button, select file with data. You can load multiple snapshots and compare them from DevTools UI.

<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388382-c264a880-8026-11e9-92c9-b455d05b89c4.png">

Tracing

thetool -o . -t tracing --recordMode recordAsMuchAsPossible --includedCategories node,v8 npm run test

To analyze: open Chrome DevTools, go to Performance tab, click load button, select file with data.

--recordMode controls how the trace buffer stores data (recordUntilFull, recordContinuously, recordAsMuchAsPossible) --includedCategories please take a look on different available categories on https://nodejs.org/api/tracing.html

E.g. you can capture V8 sampling profiler using following command:

thetool -o . -t tracing --recordMode recordAsMuchAsPossible --includedCategories v8.execute,v8.cpu_profiler,v8.cpu_profiler.hires npm run test
<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388415-1ec7c800-8027-11e9-9299-165a24dd6cdb.png">

Coverage Profiler

thetool -o . -t coverage npm run test

To analyze: in current folder create ./coverage/tmp folder and move files with data to this folder, run c8: npx c8 report. Please take a look at c8 README.md to see what output formats are supported.

<img width="1059" alt="Screen Shot 2019-05-27 at 2 10 14 AM" src="https://user-images.githubusercontent.com/426418/58388465-e2e13280-8027-11e9-9a75-e2d278d984c7.png">

Type Profiler

thetool -o . -t type npm run test

To analyze: no tool yet.