Chen Chongchen

currently this api is private https://github.com/python/cpython/blob/c77121e9f19702b1ab280299394e38e8f15c0fd3/Include/internal/pycore_unicodeobject.h#L8. I have to seek a workaround.

I cannot find these methods in final .so file. so I will copy toLower codes from python as a workaround. https://github.com/python/cpython/pull/117117

@decaz

another scenario is that, we have multi model corpus for trainning. but videos are too large, and we want to share videos or images between different datasets. so videos or...

yes, we can use >>> ds.to_table(columns={"a_sq": "a*a"}) the only difference is that, data producer can be decoupled with data consumer. otherwise when sql logic is changed, data producer has to...

> (but I believe you are also the one that suggested we introduce metadata columns? So maybe some of that schema refactor is overdue) after my pr for datafusion is...

@BubbleCal could you please help review this PR?