https-everywhere/utils/issue-format-bot/lib/parse.js at python3.8 · cleancoindev/https-everywhere

52 lines (37 loc) · 1.91 KB

// Takes in an issue body and parses it into key-value pairs, normalizes it, and returns an Object

'use strict';

const strip = require('strip-markdown'),

remark = require('remark'),

processor = remark().use(strip),

_ = require('lodash'),

domainFromPartialUrl = require('domain-from-partial-url'),

Entities = require('html-entities').AllHtmlEntities,

decode = (new Entities()).decode;

// XXX should type validation be moved to the validator module?

const validTypes = ['ruleset issue', 'new ruleset', 'code issue', 'feature request', 'other'];

// XXX the Error API in this module is pretty funky and could use a better design

module.exports = function parseDescription(body) {

const plaintext = String(processor.processSync(body)); // eslint-disable-line no-sync

// Check if there's no description at all

if (plaintext.trim().length === 0) return new Error('null description');

// Split by newlines, filter blanks, split into trimmed key-value and lowercase everything

const lines = plaintext.split('\n')

.filter(line => _.compact(line).length !== 0)

.map(line => line.split(':').map(key => key.trim()))

.map(line => [line[0].toLowerCase(), line[1]]);

const types = lines.filter(line => line[0] === 'type');

// XXX should we check for duplicates of *all* keys?

if (types.length === 0) return new Error('no type');

if (types.length > 1) return new Error('multiple types');

// `types` looks like [ [ 'Type', 'ruleset issue' ] ]

const type = types[0][1];

if (!validTypes.includes(type)) return new Error('invalid type');

// Convert to object

const normalized = _.fromPairs(lines);

// Markdown mangles full URLs into HTML entities

// (e.g. `http://example.com/` instead of

// `http://example.com/`). So we decode them again.

if (normalized.domain) normalized.domain = domainFromPartialUrl(decode(normalized.domain));

return normalized;

};

Provide feedback