When is a groupby query evaluated in RavenDB?

问题

I'm struggling with the behaviour of grouping using RavenDB and LuceneQuery.

I was always under the impression that IEnumerable was only evaluated when calling ToArray(), etc.

The following query was split into two parts for the sake of clarity.

I would not expect the query to be evaluated until after ToArray() is called on totalledBalanceList, my expectation being that the grouping is done on the server across all of the data. However, the actual result depends on the number of items stipulated in .Take(). Without the Take(1024), the results come back for the default 128 items.

I need to be able to perform the grouping across the entire dataset.

using (var session = MvcApplication.RavenSession)
{
    var computedBalanceList =
        from c in session.Advanced.LuceneQuery<Journal, Ledger_ByDateAndDebitIdAndCreditIdAndValues>()
        .Where(parameters)
        .OrderBy(c => c.DateAsString).Take(1024)
        select new LedgerBalanceDto
        {
            Account = account,
            Name = queryName,
            Debits = c.DebitId == account.Id
                         ? c.DebitValue
                         : 0,
            Credits = (c.CreditId == account.Id)
                          ? c.CreditValue
                          : 0,
            Currency = (c.DebitId == account.Id) ? c.DebitCurrency : c.CreditCurrency,
            CurrencySymbol = (c.DebitId == account.Id) ? c.DebitCurrencySymbol : c.CreditCurrencySymbol,
        };

    var totalledBalanceList =
        from balance in computedBalanceList
        group new {balance.Debits, balance.Credits} by new {balance.Currency, balance.CurrencySymbol}
        into grouping
        select new LedgerBalanceDto
        {
            Account = account,
            Currency = grouping.Key.Currency,
            CurrencySymbol = grouping.Key.CurrencySymbol,
            Debits = grouping.Sum(c => c.Debits),
            Credits = grouping.Sum(c => c.Credits),
            Name = queryName
        };

    return totalledBalanceList;

And the index:

public class Ledger_ByDateAndDebitIdAndCreditIdAndValues:AbstractIndexCreationTask<Journal>
{
    public Ledger_ByDateAndDebitIdAndCreditIdAndValues()
    {
        Map = journals => from c in journals
                          select new {c.Date,c.DateAsString, c.DebitId, c.CreditId,c.DebitValue,c.CreditValue};

        Index(x=>x.Date,FieldIndexing.Analyzed);
        Index(x=>x.DateAsString,FieldIndexing.Analyzed);
        Index(x=>x.DebitId,FieldIndexing.Analyzed);
        Index(x=>x.CreditId,FieldIndexing.Analyzed);

        Index(x=>x.DebitValue,FieldIndexing.Analyzed);
        Index(x=>x.CreditValue,FieldIndexing.Analyzed);

        Sort(x=>x.DateAsString,SortOptions.String);
    }
}

I've also rewritten the query so that the grouping happens "outside" the filter, but I get exactly the same results, namely the result depends on the Take().

var totalledBalanceList = from balance in
    from c in query
        .Where(parameters)
        .OrderBy(c => c.DateAsString)
    select new LedgerBalanceDto
    {
        Account = account,
        Name = queryName,
        Debits = c.DebitId == account.Id
                     ? c.DebitValue
                     : 0,
        Credits = (c.CreditId == account.Id)
                      ? c.CreditValue
                      : 0,
        Currency = (c.DebitId == account.Id) ? c.DebitCurrency : c.CreditCurrency,
        CurrencySymbol = (c.DebitId == account.Id) ? c.DebitCurrencySymbol : c.CreditCurrencySymbol,
    }
    group new {balance.Debits, balance.Credits} by new {balance.Currency, balance.CurrencySymbol}
    into grouping
    select new LedgerBalanceDto
    {
        Account = account,
        Currency = grouping.Key.Currency,
        CurrencySymbol = grouping.Key.CurrencySymbol,
        Debits = grouping.Sum(c => c.Debits),
        Credits = grouping.Sum(c => c.Credits),
        Name = queryName
    };
return totalledBalanceList;

Any thoughts on this would be much appreciated.

Part of the Journal class:

public class Journal
{
    public string Id { get; set; }
    public string DebitId{get;set;}
    public string CreditId{get;set;}
    public decimal? ExchangeRate { get; set; }
    public decimal CreditValue {get;set;}
    public decimal DebitValue {get;set;}
    public string DebitCurrency {get;set;}
    public string CreditCurrency {get;set;}
    public decimal Nett
    {
        get { return _nett; }
        set
        {
            _nett = value;

            CreditValue = Math.Round(Nett, 2);
            DebitValue = Math.Round(Nett * (ExchangeRate ?? 1), 2);
        }
    }
   etc ...
}

Example data IEnumerable<Journal>:

Id     DebitId       CreditId      Nett     ExchangeRate     DbCurr     CrCurr   DbVal    CrVal

1      Expense1      Bank          100      2.03             BRL        USD      203.00   100.00
2      Expense2      Bank          50       null             USD        USD      50.00     50.00
3      Bank          Client1       300      null             USD        USD      300.00   300.00
4      Stock         Bank          300      null             USD        USD      300.00   300.00

For example, when I query for the Bank, I want to be able to sum the DbVal and CrVal and calculate the balance, but in order to do so, I have to zero either one or the other (as per the query).

回答1:

You are mixing LINQ and Lucene query here, and that means they will be evaluated on the client.

You need to move a lot of this into an index and query that using session.Query, not session.Advanced.LuceneQuery.

来源：https://stackoverflow.com/questions/11963777/when-is-a-groupby-query-evaluated-in-ravendb

标签

ravendb