postgresql 中的 large object

//主要有下面两个系统表,pg_largeobject_metadata主要记录权限的,有用的只有pg_largeobjectCATALOG(pg_largeobject_metadata,2995){Oidlomowner;/* OID of the largeobject owner */#ifdef CATALOG_VARLEN/* variable-length fields start here */aclitemlomacl[1];/* access permissions */#endif} FormData_pg_largeobject_metadata;/* * Each "page" (tuple) of a large object can hold this much data * * We could set this as high as BLCKSZ less some overhead, but it seems * better to make it a smaller value, so that not as much space is used * up when a page-tuple is updated.  Note that the value is deliberately * chosen large enough to trigger the tuple toaster, so that we will * attempt to compress page tuples in-line.  (But they won't be moved off * unless the user creates a toast-table for pg_largeobject...) * * Also, it seems to be a smart move to make the page size be a power of 2, * since clients will often be written to send data in power-of-2 blocks. * This avoids unnecessary tuple updates caused by partial-page writes. */#define LOBLKSIZE  (BLCKSZ / 4)CATALOG(pg_largeobject,2613) BKI_WITHOUT_OIDS{Oidloid;/* Identifier of large object */int4pageno;/* Page number (starting from 0) *//* data has variable length, but we allow direct access; see inv_api.c */byteadata;/* Data for page (may be zero-length) */} FormData_pg_largeobject;

所有的大对象全部拆成 LOBLKSIZE  大小,放入 pg_largeobject 之中,通过 loid进行区分, 每个大对象拆分成的各个部分通过 pageno(叫元组序号更合理),进行顺序的关联起来,

pageno是通过 要写入的数据offset / LOBLKSIZE  计算的
 

pg的大对象实现的比较简单,并不适合大数据量的使用,会成为系统的瓶颈.