Datastore.select - does it work with column names that have solely integers?

[jiminikiz]

Hey dev team, we have a Datastore table like this:

As you can see, the column names are integers, i.e. IDs mapping to individual rows within another namespace > path.

Using the code found in this issue, I am able to select a column with a name like date, but not a column with a name like 4820477751066624.

transaction.select(['date']); // this works :)
transaction.select(['date','4820477751066624']); // this does not work :( 

Is it generally assumed that column names are strings with at least one letter as the starting character? Or can there be solely integer column names?

Here is my Datastore response:

{ 
    error: {
        code: 412,
        message: 'no matching index found.' 
    }
}

I have double-triple checked to make sure I run the query using an integer that is definitely a column name for the table pictured above.

Thanks in advance for your help!

[stephenplusplus]

// @pcostell

[pcostell]

Projection queries in Datastore are a little different compared to a traditional SQL projection. In particular, the projection only works if it is being run on an index (in fact, this is why we offer projection queries, so that the query only needs to perform an index scan and not the accompanied entity lookup).

In the case of transaction.select(['date','4820477751066624']);, you would need to have an index on (Date, 4820477751066624) to perform the query. However, given the automatic indexes transaction.select(['4820477751066624']); should work.

[jiminikiz]

Taking your advice and using GCloud v.0.18, here is the query:

Datastore.query({
    namespace: 'accounts',
    path: [ pathmap[ req.query.metric ] ],
    select: [n],
    filter: {
        "date >=": req.query.ds || m.startOf('month').unix(),
        "date <=": req.query.de || m.endOf('month').unix()
    }
}, function (error, stats) { 
    console.dir(error);
});

Here is the transaction object before it is sent to the Datastore servers:

{ 
    namespace: 'accounts',
    kinds: [ 'analytics' ],
    filters: 
    [ { name: 'date', op: '>=', val: 1438408800 }, { name: 'date', op: '<=', val: 1441087199 } ] ]
    orders: [],
    groupByVal: [],
    selectVal: [ '5666823336886272' ],
    autoPaginateVal: true,
    startVal: null,
    endVal: null,
    limitVal: -1,
    offsetVal: -1,
    ... [fns]
}

And here is the response:

{ 
    error: {
        code: 412,
        message: 'Precondition Failed' 
    }
}

Using the same code I can make a query and select the date column and it works. And yes I have noticed that everything is auto-indexed, so that is why I was hoping we could query by account number.

[pcostell]

You can query by date in this case because it is already part of the index you are using (due to the inequality).

In order to answer this query, you would need an index on (date, 5666823336886272).

[jiminikiz]

Okay, I get it, thank you.