v2: Implement transforms against DataFusion DataFrame, drop custom UDFs #525
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
jonmmease
commented
Oct 30, 2024
| } | ||
|
|
||
| impl VegaFusionDataset { | ||
| pub fn fingerprint(&self) -> String { | ||
| match self { | ||
| VegaFusionDataset::Table { hash, .. } => hash.to_string(), | ||
| VegaFusionDataset::DataFrame(df) => df.fingerprint().to_string(), |
There was a problem hiding this comment.
Keeping this VegaFusionDataset wrapper to hold the fingerprint, and to leave open possibility to add other plans types (e.g. DataFusion or substrait plans).
jonmmease
added a commit
that referenced
this pull request
Nov 16, 2024
…Fs (#525) * wip refactor to use DataFusion's DataFrame * Add aggregate support * Port additional transforms * Port additional transforms * Port window transform * Port fold transform * Port impute transform * Port pivot transform * port timeunit * get schema from first batch * Don't require metadata match * start porting stack * finish stack transform port * Use object-store with DataFusion to load from http * wip time functions * parse %Y-%m-%d in UTC like the browser * Update timeunit transform to use datafusion operations * remove unused UDFs * json fallback to reqwest * Fix timezone parsing * Fix selection_test * all custom spec tests passing * get image_comparison tests passing * Get all vegafusion-runtime tests passing * fix * fix * remove more udfs * remove vegafusion-datafusion-udfs, vegafusion-dataframe, and vegafusion-sql crates * fix tests * clippy fix * format * warnings / format * python test updates * Update to datafusion main * fmt * re-enable format millis test, fix substr args * Support Utf8View in json writer * fix remaining python tests * fmt * clippy fix * fmt * work around wasm-pack error * add call to update-pkg.js * remove some stale comments
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.
This PR is a big step toward the Rust goals for VegaFusion 2.0. It removes the
vegafusion-dataframeandvegafusion-sqlcrates and re-implements the data transforms against the standard DataFusion DataFrame struct. This removes several stages of indirection and complexity.This temporarily removes support for executing queries with DuckDB. The plan for re-introducing support for external SQL engines is to use the datafusion-federation crate to translate query plans to external SQL queries during evaluation. This will let us support external sql queries while still implementing transforms against the DataFusion DataFrame.
Along the way, I audited all of the custom DataFusion UDFs that were contained in the
vegafusion-datafusion-udfcrate and replaced as many as possible with standard DataFusion features. In particular, DataFusion now has very complete support for timestamps with timezones, and timezone conversion, so nearly all of the custom timestamp UDFs are now expressible with standard DataFusion expressions. This reduces our complexity and will make it easier to support external SQL execution in the future.Another change is that timestamps are now represented internally as timestamps with timezones (rather than as naive timestamps that are assumed to be in UTC). Naive timestamps are given a timezone on import.
Datecolumns are no longer converted to timestamps on import, and are handled throughout transforms as appropriate.