stricter_postcodes

[missinglink]

enforce stricter postcode parsing:

• reduce the number of country codes we load (feel free to add more)
• do not allow postcode in the first position unless its the only token.

this greatly improves autocomplete parsing of addresses with high housenumbers (>1000)

[orangejulius]

do not allow postcode in the first position unless its the only token.

This might be too strict. We often get queries like "6000, NSW, Australia" which are queries for postalcodes

[missinglink]

Yea it's unfortunate but we have way more queries such as this which need to be fixed:

      "text": "2211 SW",
      "parser": "pelias",
      "parsed_text": {
        "subject": "2211",
        "postcode": "2211"
      },

I suspect that the query you posted above will still work as expected without the postcode parse because the number will still be considered the 'subject' of the query.

[orangejulius]

Oh yeah, that should work well. 👍

[missinglink]

@orangejulius a little more leniency added :)
5ca3720

[missinglink]

Keeping track of section boundaries is super useful for things like this, it's a pity that libpostal simply discards them..