What happens?

Hello! I am testing the performance of read_parquet function , The number of data records is 143997065，Size of parquet files is about 29GB, every file has 144 columns. But in my three cases, I got different results，the difference is significant。

To Reproduce

Case 1:
29GB is a big parquet file,here is the profiling：

Case 2:
29GB parquet files are divided into 12 files，here is the profiling:

Case 3:
29GB files are divided into 128MB small files，about 268 files。

The results are different, I compared the results of the above three cases, Case 1 is similar to case 3， case 2 is slower than others,
I noticed that estimated cardinality (EC) is different, EC resulted in different query plans. could tell me How estimated cardinality (EC) is calculated when explain analyze a query?

OS:

windows10

DuckDB Version:

0.9.2

DuckDB Client:

duckdb_cli-windows-amd64.zip

Full Name:

Tom

Affiliation:

Strive

Have you tried this on the latest nightly build?

I have tested with a nightly build

Have you tried the steps to reproduce? Do they include all relevant data and configuration? Does the issue you report still appear there?

• Yes, I have