使用并行循环给每个数据行一个分数 C#

Question

I have a method which iterates through each DataRow of a DataTable then sets the final "Score" column to be a result. I am trying to figure out how to do this quicker than simply evaluating in serial. I am new to trying parallel loops and don't know if I am taking the wrong approach.

Simplified existing code:

foreach (DataRow dr in DateOptions.Rows) 
{ 
double score = evalRow(dr);
 dr["score"] = score; 
}

Using the following seems to result in error- as I am trying to modify the DataTable.

Parallel.ForEach(DateOptions.AsEnumerable(), dr =>
{
    double score = evalRow(dr);
     dr["score"] = score; 
});

Is there some way I am not thinking of to extract the result then apply the value to the appropriate column?

Answer 1

It probably depends largely on whether evalRow does anything that's not thread-safe. If the only problem is coming from modifying the DataTable, then you can likely fix it by applying Command-Query Separation: use parallel processing to figure out what to do, then drop back to serial processing to actually do it.

var rowsWithScores = DateOptions.AsEnumerable().AsParallel()
    .Select(dr => new {dr, score = evalRow(dr)})
    .ToList();
foreach(var rowWithScore in rowsWithScores)
{
    rowWithScore.dr["score"] = rowWithScore.score;
}

That said, in my experience problems like this are better solved with algorithmic fixes rather than just trying to throw parallel processing at it. If you're only dealing with thousands of items, and this is taking hours to complete, that tells me you're probably either using an algorithm with high complexity (which can probably be fixed using data structures), or doing a lot of I/O (which might lend itself to concurrent asynchronous operations). In other words, there's probably another approach that will get you orders of magnitude better performance.