Bug with RLE encoding/decoding and column chunks with both PLAIN and PLAIN_DICTIONARY data pages

ZJONSSON / parquetjs

fully asynchronous, pure JavaScript implementation of the Parquet file format

MIT License

34 stars 61 forks source link

Open garyirick-rga opened 2 years ago

garyirick-rga commented 2 years ago

Two bugs:

RLE encoding/decoding is not correct when a dictionary has > 255 entries.
A column chunk with data pages in both PLAIN and PLAIN_DICTIONARY will decode the PLAIN pages incorrectly.

garyirick-rga commented 2 years ago

I think this project is dead, but I'm hoping someone using one of the many forks finds this fix helpful.