2 59.1. Sampling Method Support Functions #
4 The TSM handler function returns a palloc'd TsmRoutine struct
5 containing pointers to the support functions described below. Most of
6 the functions are required, but some are optional, and those pointers
10 SampleScanGetSampleSize (PlannerInfo *root,
16 This function is called during planning. It must estimate the number of
17 relation pages that will be read during a sample scan, and the number
18 of tuples that will be selected by the scan. (For example, these might
19 be determined by estimating the sampling fraction, and then multiplying
20 the baserel->pages and baserel->tuples numbers by that, being sure to
21 round the results to integral values.) The paramexprs list holds the
22 expression(s) that are parameters to the TABLESAMPLE clause. It is
23 recommended to use estimate_expression_value() to try to reduce these
24 expressions to constants, if their values are needed for estimation
25 purposes; but the function must provide size estimates even if they
26 cannot be reduced, and it should not fail even if the values appear
27 invalid (remember that they're only estimates of what the run-time
28 values will be). The pages and tuples parameters are outputs.
31 InitSampleScan (SampleScanState *node,
34 Initialize for execution of a SampleScan plan node. This is called
35 during executor startup. It should perform any initialization needed
36 before processing can start. The SampleScanState node has already been
37 created, but its tsm_state field is NULL. The InitSampleScan function
38 can palloc whatever internal state data is needed by the sampling
39 method, and store a pointer to it in node->tsm_state. Information about
40 the table to scan is accessible through other fields of the
41 SampleScanState node (but note that the node->ss.ss_currentScanDesc
42 scan descriptor is not set up yet). eflags contains flag bits
43 describing the executor's operating mode for this plan node.
45 When (eflags & EXEC_FLAG_EXPLAIN_ONLY) is true, the scan will not
46 actually be performed, so this function should only do the minimum
47 required to make the node state valid for EXPLAIN and EndSampleScan.
49 This function can be omitted (set the pointer to NULL), in which case
50 BeginSampleScan must perform all initialization needed by the sampling
54 BeginSampleScan (SampleScanState *node,
59 Begin execution of a sampling scan. This is called just before the
60 first attempt to fetch a tuple, and may be called again if the scan
61 needs to be restarted. Information about the table to scan is
62 accessible through fields of the SampleScanState node (but note that
63 the node->ss.ss_currentScanDesc scan descriptor is not set up yet). The
64 params array, of length nparams, contains the values of the parameters
65 supplied in the TABLESAMPLE clause. These will have the number and
66 types specified in the sampling method's parameterTypes list, and have
67 been checked to not be null. seed contains a seed to use for any random
68 numbers generated within the sampling method; it is either a hash
69 derived from the REPEATABLE value if one was given, or the result of
72 This function may adjust the fields node->use_bulkread and
73 node->use_pagemode. If node->use_bulkread is true, which it is by
74 default, the scan will use a buffer access strategy that encourages
75 recycling buffers after use. It might be reasonable to set this to
76 false if the scan will visit only a small fraction of the table's
77 pages. If node->use_pagemode is true, which it is by default, the scan
78 will perform visibility checking in a single pass for all tuples on
79 each visited page. It might be reasonable to set this to false if the
80 scan will select only a small fraction of the tuples on each visited
81 page. That will result in fewer tuple visibility checks being
82 performed, though each one will be more expensive because it will
85 If the sampling method is marked repeatable_across_scans, it must be
86 able to select the same set of tuples during a rescan as it did
87 originally, that is a fresh call of BeginSampleScan must lead to
88 selecting the same tuples as before (if the TABLESAMPLE parameters and
92 NextSampleBlock (SampleScanState *node, BlockNumber nblocks);
94 Returns the block number of the next page to be scanned, or
95 InvalidBlockNumber if no pages remain to be scanned.
97 This function can be omitted (set the pointer to NULL), in which case
98 the core code will perform a sequential scan of the entire relation.
99 Such a scan can use synchronized scanning, so that the sampling method
100 cannot assume that the relation pages are visited in the same order on
104 NextSampleTuple (SampleScanState *node,
106 OffsetNumber maxoffset);
108 Returns the offset number of the next tuple to be sampled on the
109 specified page, or InvalidOffsetNumber if no tuples remain to be
110 sampled. maxoffset is the largest offset number in use on the page.
114 NextSampleTuple is not explicitly told which of the offset numbers in
115 the range 1 .. maxoffset actually contain valid tuples. This is not
116 normally a problem since the core code ignores requests to sample
117 missing or invisible tuples; that should not result in any bias in the
118 sample. However, if necessary, the function can use node->donetuples to
119 examine how many of the tuples it returned were valid and visible.
123 NextSampleTuple must not assume that blockno is the same page number
124 returned by the most recent NextSampleBlock call. It was returned by
125 some previous NextSampleBlock call, but the core code is allowed to
126 call NextSampleBlock in advance of actually scanning pages, so as to
127 support prefetching. It is OK to assume that once sampling of a given
128 page begins, successive NextSampleTuple calls all refer to the same
129 page until InvalidOffsetNumber is returned.
132 EndSampleScan (SampleScanState *node);
134 End the scan and release resources. It is normally not important to
135 release palloc'd memory, but any externally-visible resources should be
136 cleaned up. This function can be omitted (set the pointer to NULL) in
137 the common case where no such resources exist.