MySQL JSON_EXTRACT value of property based on criteria - mysql

Question

Suppose a JSON column called blob in a MySQL 5.7 database table called "Thing" with the following content:
[
{
"id": 1,
"value": "blue"
},
{
"id": 2,
"value": "red"
}
]
Is it possible to select all records from Thing where the blob contains an object within the array where the id is some dynamic value and the value is also some dynamic value.
E.g. "give me all Things where the blob contains an object whose id is 2 and value is 'red'"
Not sure how to form the WHERE clause below:
SET #id = 2;
SET #value1 = 'red';
SET #value2 = 'blue';
-- with equals?
SELECT *
FROM Thing
WHERE JSON_EXTRACT(blob, '$[*].id ... equals #id ... and .value') = #value1;
-- with an IN clause?
SELECT *
FROM Thing
WHERE JSON_EXTRACT(blob, '$[*].id ... equals #id ... and .value') IN (#value1, #value2);

Answer 1

They said it couldn't be done. They told me I was a fool. Lo and behold, I have done it!
Here are some helper functions to teach Hibernate how to perform JSON functions against a MySQL 5.7 backend, and an example use case. It's confusing, for sure, but it works.
Context
This contrived example is of a Person entity which can have many BioDetails, which is a question/answer type (but is more involved than that). The example within below essentially is searching within two JSON payloads, grabbing JSON values from one to build JSON paths within which to search in the other. In the end, you can pass in a complex structure of AND'd or OR'd criteria, which will be applied against a JSON blob and return only the resulting rows which match.
E.g. give my all Person entities where their age is > 30 and their favorite color is blue or orange. Given that those key/value pairs are stored in a JSON blob, you can find those matched using the example code below.
JSON-search classes and example repo
Classes below use Lombok for brevity.
Classes to allow specification of search criteria
SearchCriteriaContainer
#Data
#EqualsAndHashCode
public class SearchCriteriaContainer
{
private List<SearchCriterion> criteria;
private boolean and;
}
SearchCriterion
#Data
#EqualsAndHashCode(callSuper = true)
public class SearchCriterion extends SearchCriteriaContainer
{
private String field;
private List<String> values;
private SearchOperator operator;
private boolean not = false;
}
SearchOperator
#RequiredArgsConstructor
public enum SearchOperator
{
EQUAL("="),
LESS_THAN("<"),
LESS_THAN_OR_EQUAL("<="),
GREATER_THAN(">"),
GREATER_THAN_OR_EQUAL(">="),
LIKE("like"),
IN("in"),
IS_NULL("is null");
private final String value;
#JsonCreator
public static SearchOperator fromValue(#NotBlank String value)
{
return Stream
.of(SearchOperator.values())
.filter(o -> o.getValue().equals(value))
.findFirst()
.orElseThrow(() ->
{
String message = String.format("Could not find %s with value: %s", SearchOperator.class.getName(), value);
return new IllegalArgumentException(message);
});
}
#JsonValue
public String getValue()
{
return this.value;
}
#Override
public String toString()
{
return value;
}
}
Helper class which is used to call JSON functions
#RequiredArgsConstructor
public class CriteriaBuilderHelper
{
private final CriteriaBuilder criteriaBuilder;
public Expression<String> concat(Expression<?>... values)
{
return criteriaBuilder.function("CONCAT", String.class, values);
}
public Expression<String> substringIndex(Expression<?> value, String delimiter, int count)
{
return substringIndex(value, criteriaBuilder.literal(delimiter), criteriaBuilder.literal(count));
}
public Expression<String> substringIndex(Expression<?> value, Expression<String> delimiter, Expression<Integer> count)
{
return criteriaBuilder.function("SUBSTRING_INDEX", String.class, value, delimiter, count);
}
public Expression<String> jsonUnquote(Expression<?> jsonValue)
{
return criteriaBuilder.function("JSON_UNQUOTE", String.class, jsonValue);
}
public Expression<String> jsonExtract(Expression<?> jsonDoc, Expression<?> path)
{
return criteriaBuilder.function("JSON_EXTRACT", String.class, jsonDoc, path);
}
public Expression<String> jsonSearchOne(Expression<?> jsonDoc, Expression<?> value, Expression<?>... paths)
{
return jsonSearch(jsonDoc, "one", value, paths);
}
public Expression<String> jsonSearch(Expression<?> jsonDoc, Expression<?> value, Expression<?>... paths)
{
return jsonSearch(jsonDoc, "all", value, paths);
}
public Expression<String> jsonSearch(Expression<?> jsonDoc, String oneOrAll, Expression<?> value, Expression<?>... paths)
{
if (!"one".equals(oneOrAll) && !"all".equals(oneOrAll))
{
throw new RuntimeException("Parameter 'oneOrAll' must be 'one' or 'all', not: " + oneOrAll);
}
else
{
final var expressions = new ArrayList<>(List.of(
jsonDoc,
criteriaBuilder.literal(oneOrAll),
value,
criteriaBuilder.nullLiteral(String.class)));
if (paths != null)
{
expressions.addAll(Arrays.asList(paths));
}
return criteriaBuilder.function("JSON_SEARCH", String.class, expressions.toArray(Expression[]::new));
}
}
}
Utility to turn SearchCriteria into MySQL JSON function calls
SearchHelper
public class SearchHelper
{
private static final Pattern pathSeparatorPattern = Pattern.compile("\\.");
public static String getKeyPart(String key)
{
return pathSeparatorPattern.split(key)[0];
}
public static String getPathPart(String key)
{
final var parts = pathSeparatorPattern.split(key);
final var path = new StringBuilder();
for (var i = 1; i < parts.length; i++)
{
if (i > 1)
{
path.append(".");
}
path.append(parts[i]);
}
return path.toString();
}
public static Optional<Predicate> getCriteriaPredicate(SearchCriteriaContainer container, CriteriaBuilder cb, Path<String> bioDetailJson, Path<String> personJson)
{
final var predicates = new ArrayList<Predicate>();
if (container != null && container.getCriteria() != null && container.getCriteria().size() > 0)
{
final var h = new CriteriaBuilderHelper(cb);
container.getCriteria().forEach(ac ->
{
final var groupingOnly = ac.getField() == null && ac.getOperator() == null;
// a criterion can be used for grouping other criterion, and might not have a field/operator/value
if (!groupingOnly)
{
final var key = getKeyPart(ac.getField());
final var path = getPathPart(ac.getField());
final var bioDetailQuestionKeyPathEx = h.jsonUnquote(h.jsonSearchOne(bioDetailJson, cb.literal(key), cb.literal("$[*].key")));
final var bioDetailQuestionIdPathEx = h.concat(h.substringIndex(bioDetailQuestionKeyPathEx, ".", 1), cb.literal(".id"));
final var questionIdEx = h.jsonUnquote(h.jsonExtract(bioDetailJson, bioDetailQuestionIdPathEx));
final var answerPathEx = h.substringIndex(h.jsonUnquote(h.jsonSearchOne(personJson, questionIdEx, cb.literal("$[*].questionId"))), ".", 1);
final var answerValuePathEx = h.concat(answerPathEx, cb.literal("." + path));
final var answerValueEx = h.jsonUnquote(h.jsonExtract(personJson, answerValuePathEx));
switch (ac.getOperator())
{
case IN:
{
final var inEx = cb.in(answerValueEx);
if (ac.getValues() == null || ac.getValues().size() == 0)
{
throw new RuntimeException("No values provided for 'IN' criteria for field: " + ac.getField());
}
else
{
ac.getValues().forEach(inEx::value);
}
predicates.add(inEx);
break;
}
case IS_NULL:
{
predicates.add(cb.isNull(answerValueEx));
break;
}
default:
{
if (ac.getValues() == null || ac.getValues().size() == 0)
{
throw new RuntimeException("No values provided for '" + ac.getOperator() + "' criteria for field: " + ac.getField());
}
else
{
ac.getValues().forEach(value ->
{
final var valueEx = cb.literal(value);
switch (ac.getOperator())
{
case EQUAL:
{
predicates.add(cb.equal(answerValueEx, valueEx));
break;
}
case LESS_THAN:
{
predicates.add(cb.lessThan(answerValueEx, valueEx));
break;
}
case LESS_THAN_OR_EQUAL:
{
predicates.add(cb.lessThanOrEqualTo(answerValueEx, valueEx));
break;
}
case GREATER_THAN:
{
predicates.add(cb.greaterThan(answerValueEx, valueEx));
break;
}
case GREATER_THAN_OR_EQUAL:
{
predicates.add(cb.greaterThanOrEqualTo(answerValueEx, valueEx));
break;
}
case LIKE:
{
predicates.add(cb.like(answerValueEx, valueEx));
break;
}
default:
throw new RuntimeException("Unsupported operator during snapshot search: " + ac.getOperator());
}
});
}
}
}
}
// iterate nested criteria
getAnswerCriteriaPredicate(ac, cb, bioDetailJson, personJson).ifPresent(predicates::add);
});
return Optional.of(container.isAnd()
? cb.and(predicates.toArray(Predicate[]::new))
: cb.or(predicates.toArray(Predicate[]::new)));
}
else
{
return Optional.empty();
}
}
}
Example JPA Specification repository / search method
ExampleRepository
#Repository
public interface PersonRepository extends JpaSpecificationExecutor<Person>
{
default Page<Person> search(PersonSearchDirective directive, Pageable pageable)
{
return findAll((person, query, cb) ->
{
final var bioDetail = person.join(Person_.bioDetail);
final var bioDetailJson = bioDetail.get(BioDetailEntity_.bioDetailJson);
final var personJson = person.get(Person_.personJson);
final var predicates = new ArrayList<>();
SearchHelper
.getCriteriaPredicate(directive.getSearchCriteria(), cb, bioDetailJson, personJson)
.ifPresent(predicates::add);
return cb.and(predicates.toArray(Predicate[]::new));
}, pageable);
}
}