# Test runs

## Using --min-cluster-size

Skipping writes of single element clusters cuts clustering from ~42h to ~22h.

```
$ time zstdcat -T0 release_export_expanded.json.zst | \
    TMPDIR=/bigger/tmp python -m fuzzycat cluster --min-cluster-size 2 \
        --tmpdir /bigger/tmp -t tsandcrawler | \
    zstd -c9 > cluster_tsandcrawler_min_cluster_size_2.json.zst
...
max cluster size cut off for: 雜報その1
max cluster size cut off for: 雜録
2020-11-27 18:31:39.825 DEBUG __main__ - run_cluster: {"key_fail": 0, "key_ok":
154202433, "key_empty": 942, "key_denylist": 0, "num_clusters": 11763096}

real    1328m46.994s
user    1088m6.837s
sys     98m17.501s
```

We find 11763096 clusters, 16GB compressed (zstdcat takes about 5min,
sequential read at 50M/s).

```
$ time zstdcat -T0 cluster_tsandcrawler_min_cluster_size_2.json.zst | \
    python -m fuzzycat verify | \
    zstd -T0 -c9 > cluster_tsandcrawler_min_cluster_size_2_verify.tsv.zst
```