Performance of accessing dataSet fields using Field-names instead of indexes

Is the performance negligible?

For example,

myQuery.FieldbyName("MyField").AsString;
myQuery.Fields[0].AsString;

Cases: Table with a decent number of fields, say > 50 fields

Accessing large resultsets, say > 100,000 rows

Is the readability benefit of field names w开发者_如何学编程orth the performance decrease?

---

Here is an interesting post by François Gaillard about FieldByName performance issues.

---

The performance may not be negligible, depending on how often you access the field by name. If you use it for every field and every row you may notice a performance decrease (see for example http://www.delphifeeds.com/go/s/74559). To mantain readability yet improve performance you could:

1. Use the ['FieldName'] or FieldByName() syntax only once, and store a reference to the field in a variable.
2. Use "static" persistent field declaration, right-clicking the dataset, select Field Editor and adding needed fields. It will declare the proper TField descendant, and let you assign a name.

Also the AsXXXXX calls may be slower than using a TField descendant native Value property.

---

I have found FieldByName to be noticeable slower.

I normally access the database through an intermediate layer, that access entire records from the same table alot of times. On creation of that layer I assign the index of each field to an variable. I then use the variables for later access, to still have readable code.

ADODataSet.CommandText := 'select * from [TABLE] where 1 = 0'; //table layout
ADODataSet.Open;
ADODataSet.GetFieldNames(List);
varMyField := List.IndexOf('MyField');