< * | *** >
Others 
- reduce memory and disk usage
- need to force a value during an insert
- function sb_export
- set_text/set_sparse_text
- show table statistics (work in progress)
- allow errors in select
- sb_parallel
- fn_compute_info / fn_smart_delete (beta)
- set_dirty
- backup
- dates
- SB tags
- index vs no_index read
- countsequence
- hide data from memory
- countdistinct & high sparsity
- number format/precision
Others
reduce memory and disk usage
Use QUERY_TABLE_ALLOWED and SKIP_REFRESH_FOR_NOT_IN_MEMORY_BIG_TABLES
Use these parameters when you need to limit memory and disk usage, and when response time does not matter
need to force a value during an insert
use INSERT_FORCE_VALUE parameter
function sb_export
loop table_name(columns) function sb_export('path/to/directory','file_tag',thread_count) ...set_text/set_sparse_text
show table statistics (work in progress)
allow errors in select
Sometimes it can't be easier to ignore errors, it facilitates integration with reporting tool
Parameters: ALLOW_WHERE_ERROR, ALLOW_GROUP_BY_ERROR, ALLOW_EXP_ERROR
sb_parallel
When you use a custom C in multi thread context, you might need each thread to treat a given subset of the data
This is possible in SB, and this is called "sb_parallel mode"
If a where clause use sb_parallel lua function then the "sb_parallel mode" is triggered
For instance here under we want each thread to treat data of a single item_id, hence the sb_parallel(item_id) where clause
In some cases "sb_parallel mode" is replaced by a "mono thread mode"
This happens when pareto in used or when the index of the sb_parallel column cannot be used
fn_compute_info / fn_smart_delete (beta)
set_dirty
refresh a particular view set_dirty <the_view>backup
backup 'the_directory'dates
SB tags
select SB use_index|no_index|log_verbose|no_cache|explain
.. from ...Several SB tags can be used
index vs no_index read
index read versus no_index (aka sequential read or full scan) readcountsequence
countsequence(the_column) counts the number of value changes during reading countsequence can be an alternative to countdistinct !!! countsequence will return incorrect result if PARETO_LIMIT!=1 !!!hide data from memory
Parameters: MAX_PERF_FILTER & MAX_PERF_FILTER_FN
countdistinct & high sparsity
Parameters: SEQUENCE_COLUMNS & SEQUENCE_COLUMN_COMPANIONS & USE_INDEX_WITH_SORT_LIMIT & REPLACE_COUNTDISTINCT_BY_COUNTSEQUENCE
SEQUENCE_COLUMNS will tell SB to sort the lines according to a given column