Closed sadikovi closed 6 years ago
Files with Coverage Reduction | New Missed Lines | % | ||
---|---|---|---|---|
encodings/encoding.rs | 1 | 94.64% | ||
errors.rs | 2 | 25.0% | ||
file/reader.rs | 13 | 96.64% | ||
util/io.rs | 28 | 60.0% | ||
column/page.rs | 39 | 61.76% | ||
schema/types.rs | 72 | 90.41% | ||
file/metadata.rs | 154 | 45.82% | ||
<!-- | Total: | 309 | --> |
Totals | |
---|---|
Change from base Build 533: | -8.1% |
Covered Lines: | 10473 |
Relevant Lines: | 12023 |
Thanks @sadikovi ! Very nice work 👍 . I'll try to allocate some time to review this soon.
It would be good to check the copy
/clone
calls that I make in the code.
I think it might be worth spending some time and building the code to write Row
into Parquet, so we can test write-read roundtrip.
Hi @sadikovi . Sorry for the delay. I think the WIP PR looks great! Could you split it up into several sub-PRs and submit them for separate review?
It would be good to check the copy/clone calls that I make in the code.
Will check this in the sub-PRs.
I think it might be worth spending some time and building the code to write Row into Parquet, so we can test write-read roundtrip.
Hmm not sure if this is necessary, since we already have encoder level round-trip tests. Also, we can do round-trip test for column reader/writer - will that be enough?
Thanks. I will outline and link issues on what needs to be completed. Yes, column reading/writing should be enough.
This is WIP, to test and discuss the API and implementation.
Once design has been approved, I will split work into small chunks and submit them separately.
PR adds low-level write support with the following features:
Limitations:
None
. But for that we will need to update format first.Testing:
parquet-schema
+parquet-read
to check metadata and records + Spark.