imo you should implement double pass for the evaluator. using pest tokens in the functions etc makes it gard to implement also makes it slow and not optimizable

Yes agreed! I already started implementing it for some subset. Hope to get it finished soon.

in normal you would do use this trick. but however new phones does not include the kernel module that helps you to modify the TTL using iptables

        iptables -t mangle -A PREROUTING -i "$INTERFACE" -j TTL --ttl-inc 1

        iptables -t mangle -I POSTROUTING -o "$INTERFACE" -j TTL --ttl-inc 1

on my machine that i did the basic run, the one in the link is way more faster.

``` $ time ./duckdb_cli-linux-amd64 ./basic_batched.db -c "COPY user TO 'user.csv'" 100% (00:00:20.55 elapsed)

real 0m24.162s user 0m22.505s sys 0m1.988s ```

``` $ time ./duckdb_cli-linux-amd64 ./basic_batched.db -c "COPY user TO 'user.parquet'" 100% (00:00:17.11 elapsed)

real 0m20.970s user 0m19.347s sys 0m1.841s ```

``` $ time cargo run --bin parquet --release -- basic_batched.db user -o out.parquet Finished `release` profile [optimized] target(s) in 0.11s Running `target/release/parquet basic_batched.db user -o out.parquet` Database opened in 14.828µs

SQLite to Parquet Exporter ========================== Database: basic_batched.db Page size: 4096 bytes Text encoding: Utf8 Output: out.parquet Batch size: 10000

Exporting table: user Output file: out.parquet

   user: 100000000 rows (310.01 MB) - 5.85s (17095636 rows/sec)

Export completed successfully! ========================== Table: user Rows exported: 100000000 Time taken: 5.85s Output file: out.parquet Throughput: 17095564 rows/sec File size: 310.01 MB

real 0m6.052s user 0m10.455s sys 0m0.537s ```

``` $ time cargo run --bin csv --release -- basic_batched.db -t user -o out.csv Finished `release` profile [optimized] target(s) in 0.03s Running `target/release/csv basic_batched.db -t user -o out.csv`

real 0m6.453s user 0m5.252s sys 0m1.196s ```

Nice! Thank you