Hmm there shouldn't really be any reason to involve json here at all. There hstore file should be able to handle all parsing...

It seems the code is creating a fake array by concating [] around the value and parsing that. Don't ask me why :-).

PR with test or a test case here in the thread would be welcomed

I'll have a closer look at it later on.

Well the best fix I could come up with right now is doing the following:

    return DataTypes.ARRAY.is(options.dataType, DataTypes.HSTORE) ?
      Utils._.map(JSON.parse('[' +
        value.substring(1, value.length - 1)
          .replace('\n', '\\n')
          .replace('\t', '\\t')
          .replace('\r', '\\r') +
          ']'
        ), function (v) {
          return hstore.parse(v);
        }) : hstore.parse(value);

As far as I can see it, @seth-admittedly did the right thing by using JSON.parse.

Postgres returns the Array as {""type"=>"mobile", ""type"=>"landline""}, that's the reason why he removes the first and last char and creates the fake array. The problem is, that \n\t etc. are not escaped in HStores and as far as I can see, should not be escaped.

By escaping these values before parsing the Array we bypass that problem.

There might be a nicer solution but I'm to tired to think of any at the moment.

Testwise it's easy to insert a \n or something into https://github.com/sequelize/sequelize/blob/master/test/integration/dialects/postgres/dao.test.js#L543

@janmeier would you be so kind and point me to the part where you parse arrays? The search function through my cellphone is limited ;)
I guess the best way to resolve the issue is by using the array parsing already in place and parse the hstore afterwards - depending on the array parsing implementation used right now.

@BridgeAR Array parsing for hstore happens right at the place you already found. And yes, after reviewing this again, it seems your suggestion of escaping the line breaks would be fine

@janmeier I still think there's a better solution to this. I know that the ARRAY(HSTORE) ist parsed right at that spot, but I wanted to know where you parse things like ARRAY(TEXT/STRING/NUMBER...). If that code part does not use JSON.parse it could be used for the array parsing with HSTORES

Thanks for helping in debugging this @BridgeAR :)

This does not fix the issue, since I forgot to use the replace global.
It should be

.replace(/\n/g, '\\n')

But as we've spoken before I'll check if there's a better solution for this.

I've changed the code to use pg.types.arrayParser as we spoke about yesterday - much cleaner now :)