JSON data types are for storing JSON (JavaScript Object Notation) data, as specified inRFC 7159. Such data can also be stored astext, but the JSON data types have the advantage of enforcing that each stored value is valid according to the JSON rules. There are also assorted JSON-specific functions and operators available for data stored in these data types; seeSection 9.15.
There are two JSON data types:jsonandjsonb. They accept_almost_identical sets of values as input. The major practical difference is one of efficiency. Thejsondata type stores an exact copy of the input text, which processing functions must reparse on each execution; whilejsonbdata is stored in a decomposed binary format that makes it slightly slower to input due to added conversion overhead, but significantly faster to process, since no reparsing is needed.jsonbalso supports indexing, which can be a significant advantage.
Because thejsontype stores an exact copy of the input text, it will preserve semantically-insignificant white space between tokens, as well as the order of keys within JSON objects. Also, if a JSON object within the value contains the same key more than once, all the key/value pairs are kept. (The processing functions consider the last value as the operative one.) By contrast,jsonbdoes not preserve white space, does not preserve the order of object keys, and does not keep duplicate object keys. If duplicate keys are specified in the input, only the last value is kept.
In general, most applications should prefer to store JSON data asjsonb, unless there are quite specialized needs, such as legacy assumptions about ordering of object keys.
PostgreSQLallows only one character set encoding per database. It is therefore not possible for the JSON types to conform rigidly to the JSON specification unless the database encoding is UTF8. Attempts to directly include characters that cannot be represented in the database encoding will fail; conversely, characters that can be represented in the database encoding but not in UTF8 will be allowed.
RFC 7159 permits JSON strings to contain Unicode escape sequences denoted by\uXXXX. In the input function for thejsontype, Unicode escapes are allowed regardless of the database encoding, and are checked only for syntactic correctness (that is, that four hex digits follow\u). However, the input function forjsonbis stricter: it disallows Unicode escapes for non-ASCII characters (those aboveU+007F) unless the database encoding is UTF8. Thejsonbtype also rejects\u0000(because that cannot be represented inPostgreSQL'stexttype), and it insists that any use of Unicode surrogate pairs to designate characters outside the Unicode Basic Multilingual Plane be correct. Valid Unicode escapes are converted to the equivalent ASCII or UTF8 character for storage; this includes folding surrogate pairs into a single character.
Many of the JSON processing functions described inSection 9.15will convert Unicode escapes to regular characters, and will therefore throw the same types of errors just described even if their input is of typejsonnotjsonb. The fact that thejsoninput function does not make these checks may be considered a historical artifact, although it does allow for simple storage (without processing) of JSON Unicode escapes in a non-UTF8 database encoding. In general, it is best to avoid mixing Unicode escapes in JSON with a non-UTF8 database encoding, if possible.
When converting textual JSON input intojsonb, the primitive types described byRFC7159 are effectively mapped onto nativePostgreSQLtypes, as shown inTable 8.23. Therefore, there are some minor additional constraints on what constitutes validjsonbdata that do not apply to thejsontype, nor to JSON in the abstract, corresponding to limits on what can be represented by the underlying data type. Notably,jsonbwill reject numbers that are outside the range of thePostgreSQLnumericdata type, whilejsonwill not. Such implementation-defined restrictions are permitted byRFC7159. However, in practice such problems are far more likely to occur in other implementations, as it is common to represent JSON'snumberprimitive type as IEEE 754 double precision floating point (whichRFC7159 explicitly anticipates and allows for). When using JSON as an interchange format with such systems, the danger of losing numeric precision compared to data originally stored byPostgreSQLshould be considered.
Conversely, as noted in the table there are some minor restrictions on the input format of JSON primitive types that do not apply to the correspondingPostgreSQLtypes.
Table 8.23. JSON primitive types and correspondingPostgreSQLtypes
JSON primitive type
PostgreSQL
type
Notes
string
text
\u0000is disallowed, as are non-ASCII Unicode escapes if database encoding is not UTF8
number
numeric
NaNandinfinityvalues are disallowed
boolean
boolean
Only lowercasetrueandfalsespellings are accepted
null
(none)
SQLNULLis a different concept
The input/output syntax for the JSON data types is as specified inRFC7159.
The following are all validjson(orjsonb) expressions:
As previously stated, when a JSON value is input and then printed without any additional processing,jsonoutputs the same text that was input, whilejsonbdoes not preserve semantically-insignificant details such as whitespace. For example, note the differences here:
One semantically-insignificant detail worth noting is that injsonb, numbers will be printed according to the behavior of the underlyingnumerictype. In practice this means that numbers entered withEnotation will be printed without it, for example:
However,jsonbwill preserve trailing fractional zeroes, as seen in this example, even though those are semantically insignificant for purposes such as equality checks.
Representing data as JSON can be considerably more flexible than the traditional relational data model, which is compelling in environments where requirements are fluid. It is quite possible for both approaches to co-exist and complement each other within the same application. However, even for applications where maximal flexibility is desired, it is still recommended that JSON documents have a somewhat fixed structure. The structure is typically unenforced (though enforcing some business rules declaratively is possible), but having a predictable structure makes it easier to write queries that usefully summarize a set of“documents”(datums) in a table.
JSON data is subject to the same concurrency-control considerations as any other data type when stored in a table. Although storing large documents is practicable, keep in mind that any update acquires a row-level lock on the whole row. Consider limiting JSON documents to a manageable size in order to decrease lock contention among updating transactions. Ideally, JSON documents should each represent an atomic datum that business rules dictate cannot reasonably be further subdivided into smaller datums that could be modified independently.
jsonbContainment and ExistenceTesting_containment_is an important capability ofjsonb. There is no parallel set of facilities for thejsontype. Containment tests whether onejsonbdocument has contained within it another one. These examples return true except as noted:
The general principle is that the contained object must match the containing object as to structure and data contents, possibly after discarding some non-matching array elements or object key/value pairs from the containing object. But remember that the order of array elements is not significant when doing a containment match, and duplicate array elements are effectively considered only once.
As a special exception to the general principle that the structures must match, an array may contain a primitive value:
jsonbalso has an_existence_operator, which is a variation on the theme of containment: it tests whether a string (given as atextvalue) appears as an object key or array element at the top level of thejsonbvalue. These examples return true except as noted:
JSON objects are better suited than arrays for testing containment or existence when there are many keys or elements involved, because unlike arrays they are internally optimized for searching, and do not need to be searched linearly.
Because JSON containment is nested, an appropriate query can skip explicit selection of sub-objects. As an example, suppose that we have adoccolumn containing objects at the top level, with most objects containingtagsfields that contain arrays of sub-objects. This query finds entries in which sub-objects containing both"term":"paris"and"term":"food"appear, while ignoring any such keys outside thetagsarray:
One could accomplish the same thing with, say,
but that approach is less flexible, and often less efficient as well.
On the other hand, the JSON existence operator is not nested: it will only look for the specified key or array element at top level of the JSON value.
The various containment and existence operators, along with all other JSON operators and functions are documented inSection 9.15.
jsonbIndexingGIN indexes can be used to efficiently search for keys or key/value pairs occurring within a large number ofjsonbdocuments (datums). Two GIN“operator classes”are provided, offering different performance and flexibility trade-offs.
The default GIN operator class forjsonbsupports queries with top-level key-exists operators?,?&and?|operators and path/value-exists operator@>. (For details of the semantics that these operators implement, seeTable 9.44.) An example of creating an index with this operator class is:
The non-default GIN operator classjsonb_path_opssupports indexing the@>operator only. An example of creating an index with this operator class is:
Consider the example of a table that stores JSON documents retrieved from a third-party web service, with a documented schema definition. A typical document is:
We store these documents in a table namedapi, in ajsonbcolumn namedjdoc. If a GIN index is created on this column, queries like the following can make use of the index:
However, the index could not be used for queries like the following, because though the operator?is indexable, it is not applied directly to the indexed columnjdoc:
Still, with appropriate use of expression indexes, the above query can use an index. If querying for particular items within the"tags"key is common, defining an index like this may be worthwhile:
Now, theWHEREclausejdoc -> 'tags' ? 'qui'will be recognized as an application of the indexable operator?to the indexed expressionjdoc -> 'tags'. (More information on expression indexes can be found inSection 11.7.)
Another approach to querying is to exploit containment, for example:
A simple GIN index on thejdoccolumn can support this query. But note that such an index will store copies of every key and value in thejdoccolumn, whereas the expression index of the previous example stores only data found under thetagskey. While the simple-index approach is far more flexible (since it supports queries about any key), targeted expression indexes are likely to be smaller and faster to search than a simple index.
Although thejsonb_path_opsoperator class supports only queries with the@>operator, it has notable performance advantages over the default operator classjsonb_ops. Ajsonb_path_opsindex is usually much smaller than ajsonb_opsindex over the same data, and the specificity of searches is better, particularly when queries contain keys that appear frequently in the data. Therefore search operations typically perform better than with the default operator class.
The technical difference between ajsonb_opsand ajsonb_path_opsGIN index is that the former creates independent index items for each key and value in the data, while the latter creates index items only for each value in the data.[6]Basically, eachjsonb_path_opsindex item is a hash of the value and the key(s) leading to it; for example to index{"foo": {"bar": "baz"}}, a single index item would be created incorporating all three offoo,bar, andbazinto the hash value. Thus a containment query looking for this structure would result in an extremely specific index search; but there is no way at all to find out whetherfooappears as a key. On the other hand, ajsonb_opsindex would create three index items representingfoo,bar, andbazseparately; then to do the containment query, it would look for rows containing all three of these items. While GIN indexes can perform such an AND search fairly efficiently, it will still be less specific and slower than the equivalentjsonb_path_opssearch, especially if there are a very large number of rows containing any single one of the three index items.
A disadvantage of thejsonb_path_opsapproach is that it produces no index entries for JSON structures not containing any values, such as{"a": {}}. If a search for documents containing such a structure is requested, it will require a full-index scan, which is quite slow.jsonb_path_opsis therefore ill-suited for applications that often perform such searches.
jsonbalso supportsbtreeandhashindexes. These are usually useful only if it's important to check equality of complete JSON documents. Thebtreeordering forjsonbdatums is seldom of great interest, but for completeness it is:
Objects with equal numbers of pairs are compared in the order:
Note that object keys are compared in their storage order; in particular, since shorter keys are stored before longer keys, this can lead to results that might be unintuitive, such as:
Similarly, arrays with equal numbers of elements are compared in the order:
Primitive JSON values are compared using the same comparison rules as for the underlyingPostgreSQLdata type. Strings are compared using the default database collation.
[6]For this purpose, the term“value”includes array elements, though JSON terminology sometimes considers array elements distinct from values within objects.