[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by `from_csv/json` and `schema_of_csv/json` #27804

MaxGekk · 2020-03-05T07:43:06Z

What changes were proposed in this pull request?

In the PR, I propose:

To replace matching by Literal in ExprUtils.evalSchemaExpr() to checking foldable property of the schema expression.
To replace matching by Literal in ExprUtils.evalTypeExpr() to checking foldable property of the schema expression.
To change checking of the input parameter in the SchemaOfCsv expression, and allow foldable child expression.
To change checking of the input parameter in the SchemaOfJson expression, and allow foldable child expression.

Why are the changes needed?

This should improve Spark SQL UX for from_csv/from_json. Currently, Spark expects only literals:

spark-sql> select from_csv('1,Moscow', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', ''));
Error in query: Schema should be specified in DDL format as a string literal or output of the schema_of_csv function instead of replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '');; line 1 pos 7
spark-sql> select from_json('{"id":1, "city":"Moscow"}', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', ''));
Error in query: Schema should be specified in DDL format as a string literal or output of the schema_of_json function instead of replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '');; line 1 pos 7

and only string literals are acceptable as CSV examples by schema_of_csv/schema_of_json:

spark-sql> select schema_of_csv(concat_ws(',', 0.1, 1));
Error in query: cannot resolve 'schema_of_csv(concat_ws(',', CAST(0.1BD AS STRING), CAST(1 AS STRING)))' due to data type mismatch: The input csv should be a string literal and not null; however, got concat_ws(',', CAST(0.1BD AS STRING), CAST(1 AS STRING)).; line 1 pos 7;
'Project [unresolvedalias(schema_of_csv(concat_ws(,, cast(0.1 as string), cast(1 as string))), None)]
+- OneRowRelation
spark-sql> select schema_of_json(regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', ''));
Error in query: cannot resolve 'schema_of_json(regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', ''))' due to data type mismatch: The input json should be a string literal and not null; however, got regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', '').; line 1 pos 7;
'Project [unresolvedalias(schema_of_json(regexp_replace({"item_id": 1, "item_price": 0.1}, item_, )), None)]
+- OneRowRelation

Does this PR introduce any user-facing change?

Yes, after the changes users can pass any foldable string expression as the schema parameter to from_csv()/from_json(). For the example above:

spark-sql> select from_csv('1,Moscow', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', ''));
{"id":1,"city":"Moscow"}
spark-sql> select from_json('{"id":1, "city":"Moscow"}', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', ''));
{"id":1,"city":"Moscow"}

After change the schema_of_csv/schema_of_json functions accept foldable expressions, for example:

spark-sql> select schema_of_csv(concat_ws(',', 0.1, 1));
struct<_c0:double,_c1:int>
spark-sql> select schema_of_json(regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', ''));
struct<id:bigint,price:double>

How was this patch tested?

Added new test to CsvFunctionsSuite and to JsonFunctionsSuite.

MaxGekk · 2020-03-05T07:43:33Z

@HyukjinKwon @cloud-fan Please, review the PR.

SparkQA · 2020-03-05T08:05:02Z

Test build #119371 has finished for PR 27804 at commit cd93715.

This patch fails due to an unknown error code, -9.
This patch merges cleanly.
This patch adds no public classes.

MaxGekk · 2020-03-05T08:42:29Z

jenkins, retest this, please

HyukjinKwon · 2020-03-05T09:15:38Z

As mentioned in other PRs, I am okay. I don't feel strongly. Let me know if you guys have preferences @cloud-fan and @dongjoon-hyun.

cloud-fan · 2020-03-05T12:24:37Z

sql/core/src/test/resources/sql-tests/results/csv-functions.sql.out

@@ -24,7 +24,7 @@ select from_csv('1', 1)
 struct<>
 -- !query output
 org.apache.spark.sql.AnalysisException
-Schema should be specified in DDL format as a string literal or output of the schema_of_csv function instead of 1;; line 1 pos 7
+The expression '1' must be evaluated to a valid string.;; line 1 pos 7


valid schema string?

and as an error message, seems better to have The expression '1' is not a valid schema string.

cloud-fan · 2020-03-05T12:25:43Z

I'm fine as well, though I don't think it has a large impact.

SparkQA · 2020-03-05T13:26:21Z

Test build #119375 has finished for PR 27804 at commit cd93715.

This patch passes all tests.
This patch merges cleanly.
This patch adds no public classes.

SparkQA · 2020-03-05T19:28:23Z

Test build #119408 has finished for PR 27804 at commit 874e17b.

This patch fails Spark unit tests.
This patch merges cleanly.
This patch adds no public classes.

SparkQA · 2020-03-06T00:49:22Z

Test build #119415 has finished for PR 27804 at commit 87c2cf8.

This patch passes all tests.
This patch merges cleanly.
This patch adds no public classes.

cloud-fan · 2020-03-06T04:29:47Z

thanks, merging to master!

…dable args by `from_csv/json` and `schema_of_csv/json` ### What changes were proposed in this pull request? In the PR, I propose: 1. To replace matching by `Literal` in `ExprUtils.evalSchemaExpr()` to checking foldable property of the `schema` expression. 2. To replace matching by `Literal` in `ExprUtils.evalTypeExpr()` to checking foldable property of the `schema` expression. 3. To change checking of the input parameter in the `SchemaOfCsv` expression, and allow foldable `child` expression. 4. To change checking of the input parameter in the `SchemaOfJson` expression, and allow foldable `child` expression. ### Why are the changes needed? This should improve Spark SQL UX for `from_csv`/`from_json`. Currently, Spark expects only literals: ```sql spark-sql> select from_csv('1,Moscow', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '')); Error in query: Schema should be specified in DDL format as a string literal or output of the schema_of_csv function instead of replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '');; line 1 pos 7 spark-sql> select from_json('{"id":1, "city":"Moscow"}', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '')); Error in query: Schema should be specified in DDL format as a string literal or output of the schema_of_json function instead of replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '');; line 1 pos 7 ``` and only string literals are acceptable as CSV examples by `schema_of_csv`/`schema_of_json`: ```sql spark-sql> select schema_of_csv(concat_ws(',', 0.1, 1)); Error in query: cannot resolve 'schema_of_csv(concat_ws(',', CAST(0.1BD AS STRING), CAST(1 AS STRING)))' due to data type mismatch: The input csv should be a string literal and not null; however, got concat_ws(',', CAST(0.1BD AS STRING), CAST(1 AS STRING)).; line 1 pos 7; 'Project [unresolvedalias(schema_of_csv(concat_ws(,, cast(0.1 as string), cast(1 as string))), None)] +- OneRowRelation spark-sql> select schema_of_json(regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', '')); Error in query: cannot resolve 'schema_of_json(regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', ''))' due to data type mismatch: The input json should be a string literal and not null; however, got regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', '').; line 1 pos 7; 'Project [unresolvedalias(schema_of_json(regexp_replace({"item_id": 1, "item_price": 0.1}, item_, )), None)] +- OneRowRelation ``` ### Does this PR introduce any user-facing change? Yes, after the changes users can pass any foldable string expression as the `schema` parameter to `from_csv()/from_json()`. For the example above: ```sql spark-sql> select from_csv('1,Moscow', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '')); {"id":1,"city":"Moscow"} spark-sql> select from_json('{"id":1, "city":"Moscow"}', replace('dpt_org_id INT, dpt_org_city STRING', 'dpt_org_', '')); {"id":1,"city":"Moscow"} ``` After change the `schema_of_csv`/`schema_of_json` functions accept foldable expressions, for example: ```sql spark-sql> select schema_of_csv(concat_ws(',', 0.1, 1)); struct<_c0:double,_c1:int> spark-sql> select schema_of_json(regexp_replace('{"item_id": 1, "item_price": 0.1}', 'item_', '')); struct<id:bigint,price:double> ``` ### How was this patch tested? Added new test to `CsvFunctionsSuite` and to `JsonFunctionsSuite`. Closes apache#27804 from MaxGekk/foldable-arg-csv-json-func. Authored-by: Maxim Gekk <max.gekk@gmail.com> Signed-off-by: Wenchen Fan <wenchen@databricks.com>

MaxGekk added 15 commits March 5, 2020 09:58

Add a test

cc5a934

Support foldable schema

0495763

Check null

3b11635

Re-gen csv-functions.sql.out

b5c9f71

Add a test

109cb71

Support foldable expr

d28d019

Re-gen json-functions.sql.out

3a9cf3f

Add a test

9728724

Support foldable input

10c0bc8

Re-gen csv-functions.sql.out

2f88923

Add a test

e06da2a

Support foldable input

c0868e4

Re-gen json-functions.sql.out

35492f6

Bug fix

715ea1e

Refactoring

cd93715

cloud-fan reviewed Mar 5, 2020

View reviewed changes

Address review comments

874e17b

Fix expected err in CsvFunctionsSuite and JsonFunctionsSuite

87c2cf8

HyukjinKwon approved these changes Mar 6, 2020

View reviewed changes

cloud-fan closed this in 59f1e76 Mar 6, 2020

MaxGekk deleted the foldable-arg-csv-json-func branch June 5, 2020 19:46

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by `from_csv/json` and `schema_of_csv/json` #27804

[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by `from_csv/json` and `schema_of_csv/json` #27804

Uh oh!

MaxGekk commented Mar 5, 2020 •

edited by HyukjinKwon

Loading

Uh oh!

MaxGekk commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 5, 2020

Uh oh!

MaxGekk commented Mar 5, 2020

Uh oh!

HyukjinKwon commented Mar 5, 2020

Uh oh!

cloud-fan Mar 5, 2020

Uh oh!

cloud-fan Mar 5, 2020

Uh oh!

MaxGekk Mar 5, 2020

Uh oh!

cloud-fan commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 6, 2020

Uh oh!

cloud-fan commented Mar 6, 2020

Uh oh!

Uh oh!

[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by from_csv/json and schema_of_csv/json #27804

[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by from_csv/json and schema_of_csv/json #27804

Uh oh!

Conversation

MaxGekk commented Mar 5, 2020 • edited by HyukjinKwon Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What changes were proposed in this pull request?

Why are the changes needed?

Does this PR introduce any user-facing change?

How was this patch tested?

Uh oh!

MaxGekk commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 5, 2020

Uh oh!

MaxGekk commented Mar 5, 2020

Uh oh!

HyukjinKwon commented Mar 5, 2020

Uh oh!

cloud-fan Mar 5, 2020

Choose a reason for hiding this comment

Uh oh!

cloud-fan Mar 5, 2020

Choose a reason for hiding this comment

Uh oh!

MaxGekk Mar 5, 2020

Choose a reason for hiding this comment

Uh oh!

cloud-fan commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 5, 2020

Uh oh!

SparkQA commented Mar 6, 2020

Uh oh!

cloud-fan commented Mar 6, 2020

Uh oh!

Uh oh!

[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by `from_csv/json` and `schema_of_csv/json` #27804

[SPARK-31020][SPARK-31023][SPARK-31025][SPARK-31044][SQL] Support foldable args by `from_csv/json` and `schema_of_csv/json` #27804

MaxGekk commented Mar 5, 2020 •

edited by HyukjinKwon

Loading