-
Notifications
You must be signed in to change notification settings - Fork 1
JSON Serialization
Initially with version 1 of the perj logger the JavaScript objects were converted into JSON strings (serialized) by using the fast-safe-stringify module.
It was discovered soon after the initial version 1 release that the fast-safe-stringify module had issues recursively parsing some browser specific objects such as Vue.js. It was also made apparent that because of fast-safe-stringify, using the passThrough option was literally passing the object references though to the write function. If a custom write function changed an object it would cause side effects.
With these issues in mind a new internal module was created called notation-copy.
The notation-copy module is an internal module to the perj project. It offers major improvements in the parsing of JavaScript objects both in the Node.js and Browser environments.
The notation-copy module supports the following features:
- Fast object parsing.
- Supports the same API as
Object.assign. - Deep copies JavaScript objects in preparation for JSON stringification.
- Converts
Errorobjects properties into enumerable properties. - Replaces circular references with the
[Circular]string token. - Removes non-JSON related constructs such as Functions.
- Follows the JSON.stringify MDN reference documentation for JavaScript type support.
By passing an object into the notation-copy module the returned resulting object has no references to the original object and is void of non-JSON constructs.
The notation-copy module is used to copy objects passed into the perj log methods. These sanitized objects are then stringified with a simple JSON.stringify call to provide the logger output. If the passThrough option is used with a custom write function, the object passed into the custom write function is clear of circular references and non-JSON related constructs.
Despite copying the content from logged objects, perj using the notation-copy module remains extremely fast when compared to pino. See the Performance document for more detail.