Like read-package-json, but faster and more accepting of "missing" data.

This is only suitable for reading package.json files in a node_modules tree, since it doesn't do the various cleanups, normalization, and warnings that are beneficial at the root level in a package being published.

const rpj = require('read-package-json-fast') // typical promisey type API rpj('/path/to/package.json') .then(data => ...) .catch(er => ...) // or just normalize a package manifest const normalized = rpj.normalize(packageJsonObject)

Errors raised from parsing will use json-parse-even-better-errors, so they'll be of type JSONParseError and have a code: 'EJSONPARSE' property. Errors will also always have a path member referring to the path originally passed into the function.

To preserve indentation when the file is saved back to disk, use data[Symbol.for('indent')] as the third argument to JSON.stringify, and if you want to preserve windows \r\n newlines, replace the \n chars in the string with data[Symbol.for('newline')].

For example:

const data = await readPackageJsonFast('./package.json') const indent = Symbol.for('indent') const newline = Symbol.for('newline') // .. do some stuff to the data .. const string = JSON.stringify(data, null, data[indent]) + '\n' const eolFixed = data[newline] === '\n' ? string : string.replace(/\n/g, data[newline]) await writeFile('./package.json', eolFixed)

Indentation is determined by looking at the whitespace between the initial { and the first " that follows it. If you have lots of weird inconsistent indentation, then it won't track that or give you any way to preserve it. Whether this is a bug or a feature is debatable ;)

• Parse JSON
• Normalize bundledDependencies/bundleDependencies naming to just bundleDependencies (without the extra d)
• Handle true, false, or object values passed to bundleDependencies
• Normalize funding: <string> to funding: { url: <string> }
• Remove any scripts members that are not a string value.
• Normalize a string bin member to { [name]: bin }.
• Fold optionalDependencies into dependencies.
• Set the _id property if name and version are set. (This is load-bearing in a few places within the npm CLI.)

• Warn about invalid/missing name, version, repository, etc.
• Extract a description from the README.md file, or attach the readme to the parsed data object.
• Read the HEAD value out of the .git folder.
• Warn about potentially typo'ed scripts (eg, tset instead of test)
• Check to make sure that all the files in the files field exist and are valid files.
• Fix bundleDependencies that are not listed in dependencies.
• Fix dependencies fields that are not strictly objects of string values.
• Anything involving the directories field (ie, bins, mans, and so on).