Merge pull request #270 from moj-analytical-services/sql_dialect_fix

fix sql dialect bug

CHANGELOG.md

@@ -9,6 +9,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+SQLGlot produces SQL in the wrong dialect in some instances. See [issue 269](https://github.com/moj-analytical-services/splink/issues/269)
+
+## 2.1.3
+
+### Fixed
+
 Versions of sqlglot before 1.18.0 could not parse the spark sql `left` and `right` functions - see [here](https://github.com/tobymao/sqlglot/issues/58).
 
 As a result, if `left` or `right` are used in the user's custom sql case expressions, Splink will produce an error.

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "splink"
-version = "2.1.3"
+version = "2.1.4"
 description = "Implementation of Fellegi-Sunter's canonical model of record linkage in Apache Spark, including EM algorithm to estimate parameters"
 authors = ["Robin Linacre <[email protected]>", "Sam Lindsay", "Theodore Manassis"]
 license = "MIT"

splink/parse_case_statement.py

@@ -63,11 +63,9 @@ def _parse_top_level_case_statement_from_sql(top_level_case_tree):
     for i in ifs:
         lit = i.args["true"].sql()
 
-        # sql = i.args["this"].sql()
-        sql = i.sql()
+        sql = i.args["this"].sql(dialect="spark")
+        sql = f"WHEN {sql} THEN {lit}".format(sql=sql, lit=lit)
 
-        sql = re.sub(r"^CASE ", "", sql)
-        sql = re.sub(r" END$", "", sql)
         parsed_case_expr.append({"sql_expr": sql, "label": f"level_{lit}"})
 
     if top_level_case_tree.args.get("default") is not None: