Should this package try to parse like PostgreSQL, or just what PostgreSQL produces?

[charmander]

Take the boolean parser, for example:

node-pg-types/lib/textParsers.js

Lines 14 to 23 in fbe5b0e

	function parseBool (value) {
	if (value === null) return value
	return value === 'TRUE' ||
	value === 't' ||
	value === 'true' ||
	value === 'y' ||
	value === 'yes' ||
	value === 'on' ||
	value === '1'
	}

It checks for 'TRUE', 't', 'true', 'y', 'yes', 'on', and '1', and assumes anything else is false. This seems like overkill for interpreting fields from the server to me, since PostgreSQL only sends 't' and 'f' as text representations… but it’s not the right way to parse booleans in general the way PostgreSQL would either, since the latter

• is case-insensitive ('True' is valid)
• ignores surrounding whitespace (E' yes\t\n\f\r' is valid)
• allows prefixes of words ('of', 'tru', 'fals' are valid)

In other words, there are strings PostgreSQL will parse as true that parseBool will parse as false with no warning.

So is 't' → true, 'f' → false, otherwise → throw a good direction? I’ll take a look at how it affects the performance of pg at some point.

[bendrucker]

That seems good to me!

[sehrope]

+1