You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2022/02/18 23:14:07 UTC
[GitHub] [spark] andygrove opened a new pull request #35573: [WIP] [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
andygrove opened a new pull request #35573:
URL: https://github.com/apache/spark/pull/35573
Signed-off-by: Andy Grove <an...@gmail.com>
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a faster review.
7. If you want to add a new configuration, please read the guideline first for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
8. If you want to add or modify an error type or message, please read the guideline first in
'core/src/main/resources/error/README.md'.
-->
### What changes were proposed in this pull request?
When parsing JSON unquoted `NaN` and `Infinity`values for floating-point columns we get the expected behavior as shown below where valid values are returned when the parsing option `allowNonNumericNumbers` is enabled and `null` otherwise.
| Value | allowNonNumericNumbers=true | allowNonNumericNumbers=false |
| --------- | --------------------------- | ---------------------------- |
| NaN | Double.NaN | null |
| +INF | Double.PositiveInfinity | null |
| +Infinity | Double.PositiveInfinity | null |
| Infinity | Double.PositiveInfinity | null |
| -INF | Double.NegativeInfinity | null |
| -Infinity | Double.NegativeInfinity | null |
However, when these values are quoted we get the following unexpected behavior due to a different code path being used that is inconsistent with Jackson's parsing and that ignores the `allowNonNumericNumbers` parser option.
| Value | allowNonNumericNumbers=true | allowNonNumericNumbers=false |
| ----------- | --------------------------- | ---------------------------- |
| "NaN" | Double.NaN | Double.NaN |
| "+INF" | null | null |
| "+Infinity" | null | null |
| "Infinity" | Double.PositiveInfinity | Double.PositiveInfinity |
| "-INF" | null | null |
| "-Infinity" | Double.NegativeInfinity | Double.NegativeInfinity |
This PR updates the code path that handles quoted non-numeric numbers to make it consistent with the path that handles the unquoted values.
### Why are the changes needed?
<!--
Please clarify why the changes are needed. For instance,
1. If you propose a new API, clarify the use case for a new API.
2. If you fix a bug, you can clarify why it is a bug.
-->
The current behavior does not match the documented behavior in https://spark.apache.org/docs/latest/sql-data-sources-json.html
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such as the documentation fix.
If yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description and/or an example to show the behavior difference if possible.
If possible, please also clarify if this is a user-facing change compared to the released Spark versions or within the unreleased branches such as master.
If no, write 'No'.
-->
Yes, parsing of quoted `NaN` and `Infinity` values will now be consistent with the unquoted versions.
### How was this patch tested?
<!--
If tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible.
If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future.
If tests were not added, please describe why they were not added and/or why it was difficult to add.
If benchmark tests were added, please run the benchmarks in GitHub Actions for the consistent environment, and the instructions could accord to: https://spark.apache.org/developer-tools.html#github-workflow-benchmarks.
-->
Unit tests are updated.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] srowen commented on pull request #35573: [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
srowen commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1046116321
I see, so the most consistent behavior would be to not allow `"NaN"`, whether allowing non-numeric inputs to be parsed as numbers or not?
So a raw literal like `NaN` is not handled by this section of code in Spark, but by Jackson?
Well in any event, I suppose it's a little better to be permissive in this narrow case, to achieve a little more consistency, rather than 'break' existing functionality.
`allowNonNumericNumbers` is on by default, so adding the additional constraint here wouldn't affect most paths. If someone had explicitly disallowed it, now Spark would parse things like `"NaN"` as NaN. It could be worth a note in core-migration-guide.md for good measure.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] andygrove commented on pull request #35573: [WIP] [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
andygrove commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1046114075
> Dumb questions as I just don't know this code path -- would we parse `"1.0"` as a number or reject it for being a string? and you're saying that Jackson would parse `"NaN"` as NaN and not a string? It feels weird to parse strings that aren't even numbers as numbers, even when the option to do so is set, but if that is consistent with other behavior I could see it.
Jackson will parse valid JSON numbers as either integers or floats. It will parse strings as strings and does not try to interpret them. Jackson has special behavior for tokens such as `NaN` and `Infinity` which are not valid JSON values (JSON does not support non-numeric numbers at all). Jackson will not attempt to interpret `"NaN"` and would just return a string.
For some reason, the code we have today in Spark's `JacksonParser` will look at any returned strings and will parse them if they are `NaN`, `Infinity` or `-Infinity` (but not the other special values that Jackson supports, such as `+INF`). It will not attempt to parse strings that contain actual numeric values like `"1.0"`. This secondary parsing seems like a bad idea overall IMO. Users could get similar behavior by specifying that the column is `StringType` when reading and then performing an explicit `CAST` to a floating-point type and it would support reading numbers and not-numbers regardless of whether they are JSON strings or JSON numbers.
If we have the option of removing this code (or putting it behind a configuration option) then I would be in favor of that. If not, then I would like it to support the same set of not-a-number strings as Jackson and I would like it to respect the `allowNonNumericNumbers` option.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] andygrove commented on pull request #35573: [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
andygrove commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1046116794
> I see, so the most consistent behavior would be to not allow `"NaN"`, whether allowing non-numeric inputs to be parsed as numbers or not?
Yes.
> So a raw literal like `NaN` is not handled by this section of code in Spark, but by Jackson?
That is correct.
> Well in any event, I suppose it's a little better to be permissive in this narrow case, to achieve a little more consistency, rather than 'break' existing functionality.
Yes, I agree.
> `allowNonNumericNumbers` is on by default, so adding the additional constraint here wouldn't affect most paths. If someone had explicitly disallowed it, now Spark would parse things like `"NaN"` as NaN. It could be worth a note in core-migration-guide.md for good measure.
Ok, I will add a note to the migration guide in this PR later today.
Thanks for the review so far.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] AmplabJenkins commented on pull request #35573: [WIP] [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
AmplabJenkins commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1046083341
Can one of the admins verify this patch?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] srowen commented on pull request #35573: [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
srowen commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1047867287
Merged to master
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] HyukjinKwon commented on pull request #35573: [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
HyukjinKwon commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1047335476
i'm fine with this change.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] srowen closed pull request #35573: [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
srowen closed pull request #35573:
URL: https://github.com/apache/spark/pull/35573
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] [spark] srowen commented on pull request #35573: [WIP] [SPARK-38060] [SQL] Respect allowNonNumericNumbers when parsing quoted NaN and Infinity values in JSON reader
Posted by GitBox <gi...@apache.org>.
srowen commented on pull request #35573:
URL: https://github.com/apache/spark/pull/35573#issuecomment-1046102850
Dumb questions as I just don't know this code path -- would we parse `"1.0"` as a number or reject it for being a string? and you're saying that Jackson would parse `"NaN"` as NaN and not a string? It feels weird to parse strings that aren't even numbers as numbers, even when the option to do so is set, but if that is consistent with other behavior I could see it.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org