dynamic reading speed

[JoshClose]

Dynamic reading is really slow. Look into performance.

[shibayan]

ExpandoObject locks every time during an add operation. As a result, dynamic reading seems to be very slow.

To solve this problem will need an implementation that does not use ExpandoObject. (Like SqlMapper.DapperRow class in Dapper)

I think there are two implementation plans.

1. Implement inherit Dictionary<string, object> that implements IDynamicMetaObjectProvider.

• It is easy to implement, but it is not memory efficient.

2. Create a class that implements IDictionary<string, object> and IDynamicMetaObjectProvider specialized for CSV

• Improve memory efficiency by sharing CSV header information.

I write and test the code to improve this problem. Thanks.

[superrnovae]

Itterating with foreach over IEnumerable<dynamic> is extremely slow and takes around 5 minutes for 50 thousand records.

[shibayan]

@superrnovae An auxiliary library has been created for fast handling. I believe it will help. https://github.com/shibayan/CsvHelper.FastDynamic

[JoshClose]

@shibayan So this makes using dynamic faster than using types?

[shibayan]

@JoshClose No, it will not be faster than accessing with types because of the overhead present. However, performance should be significantly improved compared to dynamic using ExpandoObject.

[JoshClose]

This will be in the next release, 32.0.0.