Better hint sniffing #57
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
tests/unit/resources/hint_sniffing/delimited-tsv-with-header.json
Outdated
Show resolved
Hide resolved
vinceatbluelabs
commented
May 13, 2020
|
|
||
|
|
||
| logger = logging.getLogger(__name__) | ||
|
|
||
| HINT_INFERENCE_SAMPLING_SIZE_BYTES = 1024 |
There was a problem hiding this comment.
Let me know if you think this magic number feels important enough to add in as a ProcessingInstruction. My gut is that the default value is small enough not to matter performance-wise and big enough that we're not likely to get better information out of the hint sniffers by tweaking it higher.
This could potentially be something we offer config via the mechanism in #47 - though if it does make sense as a parameter, I'm not sure it makes sense as a system-wide parameter...
|
(whoops, need to figure out the integration test failure!) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The Pandas hint inference is actually not doing that much for us. The following things aren't inferred at all by it:
This is confusing, as our current code nominally pulls that info from a pandas object - but in practice, those things are either pulled from what we originally passed in to read_csv() or set to default values - https://github.com/pandas-dev/pandas/blob/e9b019b653d37146f9095bb0522525b3a8d9e386/pandas/io/parsers.py#L2253
It turns out that Python itself ships with an csv.Sniffer class that can do a lot of this. There are some limitations - it only works with some common record terminators and it only operates on decompressed files, but those are things we can work around.
The net effect is a lot better hint sniffing overall, which shows in the component test results here (sample CSV need fewer initial hints to be able to understand the files).