PocoData.QueryColumns should ignore read-only properties

[asherber]

public class Foo
{
    public int ID { get; set; }
    public DateTime DOB { get; set; }
    public int Month => DOB.Month;
}

using (var db = new Database(...))
{
    db.Query<Foo>().Dump();
    // SqlException: Invalid column name 'Month'.
}

I know that there's an Ignore attribute for this sort of thing, but it seems like PP should also ignore read-only properties when auto-generating a SELECT clause.

I think this could be done by changing the PocoData constructor like so, to check for read-only properties:

QueryColumns = (
    from c in Columns
    where (!c.Value.ResultColumn || c.Value.AutoSelectedResultColumn) && c.Value.PropertyInfo.CanWrite
    select c.Key
).ToArray();

I'm happy to work on a PR if this sounds right.

[pleb]

If I'm not mistaken, that's a Linqpad query. .Dump(). I too am a Linqpad addict 😄

Anyway, getting back on topic. Yes, absolutely. Seems logical.

[asherber]

To be consistent, I think this also needs to be handled within PocoData.GetFactory() so that we don't try to map from a table column to a read-only property which happens to have the same name (probably quite an edge case, but I can imagine a valid scenario).

And looking through the code for other quirks, I found something slightly related, which is that Ignore fields are only ignored during querying, not during update/delete; I assumed from the name that they would always be ignored. Should they be?

Which leads me to the observation that write-only fields should probably be excluded from inserts/updates, since trying to get their values will throw an exception.

I'm thinking of situations in which a POCO does not perfectly correspond to a table row and a user might want more control over how and when things get mapped back and forth. Would love to hear some discussion on this topic before proceeding.