[PostgreSQL tutorial] large object in PostgreSQL

1. // There are mainly the following two system tables. pg_largeobject_metadata records the permissions mainly. Only pg_largeobject is useful.
5. CATALOG (FIG, 2995)
6. {
7. Oid lomowner;/* OID of the largeobject owner */
8. # Ifdef CATALOG_VARLEN/* variable-length fields start here */
9. Aclitem lomacl [1];/* Access permissions */
10. # Endif
11. } FormData_pg_largeobject_metadata;
13. /* 
14. * Each "page" (tuple) of a large object can hold this much data 
15. * 
16. * We cocould set this as high as BLCKSZ less some overhead, but it seems 
17. * Better to make it a smaller value, so that not as much space is used 
18. * Up when a page-tuple is updated. Note that the value is deliberately 
19. * Chosen large enough to trigger the tuple toaster, so that we will 
20. * Attempt to compress page tuples in-line. (But they won't be moved off 
21. * Unless the user creates a toast-table for pg_largeobject ...) 
22. * 
23. * Also, it seems to be a smart move to make the page size be a power of 2, 
24. * Since clients will often be written to send data in power-of-2 blocks. 
25. * This avoids unnecessary tuple updates caused by partial-page writes. 
26. */
27. # Define LOBLKSIZE (BLCKSZ/4)
29. CATALOG (pg_largeobject, 2613) BKI_WITHOUT_OIDS
30. {
31. Oid loid;/* Identifier of large object */
32. Int4 pageno;/* Page number (starting from 0 )*/
33. /* Data has variable length, but we allow direct access; see inv_api.c */
34. Bytea data;/* Data for page (may be zero-length )*/
35. } FormData_pg_largeobject;

All the large objects are split into LOBLKSIZE values and placed into pg_largeobject. They are differentiated by loid. Each part of a large object is divided by pageno (the sequence number of the tuples is more reasonable ), in order,

Pageno is calculated based on the data offset/LOBLKSIZE to be written.

Pg's large object implementation is relatively simple and is not suitable for the use of large data volumes and will become a bottleneck of the system.