sqlite index to optimize NOT queries - sqlite

When I make an index in sqlite, it optimizes queries for that key. But sometimes I want to find records that exclude certain values of that key, and my index doesn't seem to be used to optimize those, at all. I can't figure out how to write an index to optimize a query that matches NOT on a key.
This is an example of what I'm talking about:
CREATE TABLE place (
id INTEGER PRIMARY KEY,
thing INTEGER NOT NULL,
thekey INTEGER NOT NULL,
unique(thing,thekey));
CREATE INDEX bythekeys ON place(thekey);
EXPLAIN QUERY PLAN SELECT thing FROM place WHERE NOT thekey = ?;
-- => 0 0 0 SCAN TABLE place
EXPLAIN QUERY PLAN SELECT thing FROM place WHERE thekey = ?;
-- => 0 0 0 SEARCH TABLE place USING INDEX bythekeys (thekey=?)
CREATE INDEX bythenotkeys ON ........?
The "bythekeys" index optimizes queries that look up records by that key, unless the query uses the logical negation of that lookup. Whether using NOT, or != it doesn't seem to make a difference. It always just scans the whole table without using any index. Do I make, like... a partial index or something? How do I optimize NOT queries?

The database assumes that a column contains many different values.
So using a != filter is estimated to not reduce the number of results enough to make the additional index lookups worthwhile.
If you actually have very few values, rewrite the query so that it uses equality (=), or a range (<, >).

Related

When to create multi-column indices in SQLite?

Assume I have a table in an SQLite database:
CREATE TABLE orders (
id INTEGER PRIMARY KEY,
price INTEGER NOT NULL,
updateTime INTEGER NOT NULL,
) [WITHOUT ROWID];
what indices should I create to optimize the following query:
SELECT * FROM orders WHERE price > ? ORDER BY updateTime DESC;
Do I create two indices:
CREATE INDEX i_1 ON orders(price);
CREATE INDEX i_2 ON orders(updateTime);
or one complex index?
CREATE INDEX i_3 ON orders(price, updateTime);
What can be query time complexity?
From The SQLite Query Optimizer Overview/WHERE Clause Analysis:
If an index is created using a statement like this:
CREATE INDEX idx_ex1 ON ex1(a,b,c,d,e,...,y,z);
Then the index might
be used if the initial columns of the index (columns a, b, and so
forth) appear in WHERE clause terms. The initial columns of the index
must be used with the = or IN or IS operators. The right-most column
that is used can employ inequalities.
As explained also in The SQLite Query Optimizer Overview/The Skip-Scan Optimization with an example:
Because the left-most column of the index does not appear in the WHERE
clause of the query, one is tempted to conclude that the index is not
usable here. However, SQLite is able to use the index.
This means than if you create an index like:
CREATE INDEX idx_orders ON orders(updateTime, price);
it might be used to optimize the WHERE clause even though updateTime does not appear there.
Also, from The SQLite Query Optimizer Overview/ORDER BY Optimizations:
SQLite attempts to use an index to satisfy the ORDER BY clause of a
query when possible. When faced with the choice of using an index to
satisfy WHERE clause constraints or satisfying an ORDER BY clause,
SQLite does the same cost analysis described above and chooses the
index that it believes will result in the fastest answer.
Since updateTime is defined first in the composite index, the index may also be used to optimize the ORDER BY clause.

is it good idea to use a binary attribute for GSI indexing in dynamo DB?

I have one attribute in my DynamoDB table which will take binary values success and failure.
can I do GSI indexing on this attribute if i have to fetch/query either success or failure records from this table?
or should i make two different table table for success and failure scenarios?
if should not do indexing on binary attribute,
what are the problems with GSI indexing of binary attribute?
how it will impact performance of query operation?
It sounds like you perhaps mean boolean (true/false) rather than binary. You cannot create a GSI on a boolean attribute in DynamoDB but you can on a string, number or binary attribute (which is different to boolean), so you can consider 1 / 0 or “accept” / “fail” for your logical boolean.
You might consider making this a sparse index if you only want to query one side of your index. So if you only want to query when there is a true (or “accept” or 1 or anything really) then when it is not true, delete the attribute rather than set it to “failure” or 0 etc. This makes queries far more performant as the index is smaller, but the limitation is you can no longer query the “failure” / false / 0 cases.
To answer your questions:
1) you can’t create an index on a boolean, use a string or number (or binary, but probably you want string or number)
2) if you only need to query one side of the boolean (e.g. “accept” but never “failure”) you can improve the performance by creating a sparse index

How do I create an expression-based index using the instr() function?

With Sqlite3, I am trying to do a query like:
select *
from data
where instr(filepath,'.txt') != 0
And I want to index this query to speed it up.
I tried to create an index like:
create index data_instr_filepath
on data(instr(filepath,'.txt'));
However, "explain query plan" still shows that I'm doing a table scan.
Is this doable in sqlite? The examples I have found for doing expression-based indexes seems to be limited to the length function and multiplying two columns together.
UPDATE:
Thanks to Mike's answer, I refactored my query to not use inequalities and was able to create an index that hits it. Below are my indexes that I ended up using:
create index data_instr_filepath_txt on data(instr(filepath,'.txt'));
create index data_instr_filepath_substr on data(substr(filepath,0,instr(filepath,'.')));
The reason is that an index will likely not be used for an inequality as per :-
Similarly, index columns will not normally be used (for indexing
purposes) if they are to the right of a column that is constrained
only by inequalities. The SQLite Query Optimizer Overview
You are able to try forcing the use of an index by using INDEXED BY. However, this will not work in your situation because of the above flagging the index as not being usable. (the query will still work)
e.g.
EXPLAIN QUERY PLAN
SELECT * FROM data INDEXED BY data_instr_filepath
WHERE instr(filepath,'.txt') != 0
results in :-
no query solution
Time: 0s

how can I get faster FTS4 query results ordered by a field in another table?

Background
I'm implementing full-text search over a body of email messages stored in SQLite, making use of its fantastic built-in FTS4 engine. I'm getting some rather poor query performance, although not exactly where I would expect. Let's take a look.
Representative schema
I'll give some simplified examples of the code in question, with links to the full code where applicable.
We've got a MessageTable that stores the data about an email message (full version spread out over several files here, here, and here):
CREATE TABLE MessageTable (
id INTEGER PRIMARY KEY,
internaldate_time_t INTEGER
);
CREATE INDEX MessageTableInternalDateTimeTIndex
ON MessageTable(internaldate_time_t);
The searchable text is added to an FTS4 table named MessageSearchTable (full version here):
CREATE VIRTUAL TABLE MessageSearchTable USING fts4(
id INTEGER PRIMARY KEY,
body
);
The id in the search table acts as a foreign key to the message table.
I'll leave it as an exercise for the reader to insert data into these tables (I certainly can't give out my private email). I have just under 26k records in each table.
Problem query
When we retrieve search results, we need them to be ordered descending by internaldate_time_t so we can pluck out only the most recent few results. Here's an example search query (full version here):
SELECT id
FROM MessageSearchTable
JOIN MessageTable USING (id)
WHERE MessageSearchTable MATCH 'a'
ORDER BY internaldate_time_t DESC
LIMIT 10 OFFSET 0
On my machine, with my email, that runs in about 150 milliseconds, as measured via:
time sqlite3 test.db <<<"..." > /dev/null
150 milliseconds is no beast of a query, but for a simple FTS lookup and indexed order, it's sluggish. If I omit the ORDER BY, it completes in 10 milliseconds, for example. Also keep in mind that the actual query has one more sub-select, so there's a little more work going on in general: the full version of the query runs in about 600 milliseconds, which is into beast territory, and omitting the ORDER BY in that case shaves 500 milliseconds off the time.
If I turn on stats inside sqlite3 and run the query, I notice the line:
Sort Operations: 1
If my interpretation of the docs about those stats is correct, it looks like the query is completely skipping using the MessageTableInternalDateTimeTIndex. The full version of the query also has the line:
Fullscan Steps: 25824
Sounds like it's walking the table somewhere, but let's ignore that for now.
What I've discovered
So let's work on optimizing that a little bit. I can rearrange the query into a sub-select and force SQLite to use our index with the INDEXED BY extension:
SELECT id
FROM MessageTable
INDEXED BY MessageTableInternalDateTimeTIndex
WHERE id IN (
SELECT id
FROM MessageSearchTable
WHERE MessageSearchTable MATCH 'a'
)
ORDER BY internaldate_time_t DESC
LIMIT 10 OFFSET 0
Lo and behold, the running time has dropped to around 100 milliseconds (300 milliseconds in the full version of the query, a 50% reduction in running time), and there are no sort operations reported. Note that with just reorganizing the query like this but not forcing the index with INDEXED BY, there's still a sort operation (though we've still shaved off a few milliseconds oddly enough), so it appears that SQLite is indeed ignoring our index unless we force it.
I've also tried some other things to see if they'd make a difference, but they didn't:
Explicitly making the index DESC as described here, with and without INDEXED BY
Explicitly adding the id column in the index, with and without internaldate_time_t ordered DESC, with and without INDEXED BY
Probably several other things I can't remember at this moment
Questions
100 milliseconds here still seems awfully slow for what seems like it should be a simple FTS lookup and indexed order.
What's going on here? Why is it ignoring the obvious index unless you force its hand?
Am I hitting some limitation with combining data from virtual and regular tables?
Why is it still so relatively slow, and is there anything else I can do to get FTS matches ordered by a field in another table?
Thanks!
An index is useful for looking up a table row based on the value of the indexed column.
Once a table row is found, indexes are no longer useful because it is not efficient to look up a table row in an index by any other criterium.
An implication of this is that it is not possible to use more than one index for each table accessed in a query.
Also see the documentation: Query Planning, Query Optimizer.
Your first query has the following EXPLAIN QUERY PLAN output:
0 0 0 SCAN TABLE MessageSearchTable VIRTUAL TABLE INDEX 4: (~0 rows)
0 1 1 SEARCH TABLE MessageTable USING INTEGER PRIMARY KEY (rowid=?) (~1 rows)
0 0 0 USE TEMP B-TREE FOR ORDER BY
What happens is that
the FTS index is used to find all matching MessageSearchTable rows;
for each row found in 1., the MessageTable primary key index is used to find the matching row;
all rows found in 2. are sorted with a temporary table;
the first 10 rows are returned.
Your second query has the following EXPLAIN QUERY PLAN output:
0 0 0 SCAN TABLE MessageTable USING COVERING INDEX MessageTableInternalDateTimeTIndex (~100000 rows)
0 0 0 EXECUTE LIST SUBQUERY 1
1 0 0 SCAN TABLE MessageSearchTable VIRTUAL TABLE INDEX 4: (~0 rows)
What happens is that
the FTS index is used to find all matching MessageSearchTable rows;
SQLite goes through all entries in the MessageTableInternalDateTimeTIndex in the index order, and returns a row when the id value is one of the values found in step 1.
SQLite stops after the tenth such row.
In this query, it is possible to use the index for (implied) sorting, but only because no other index is used for looking up rows in this table.
Using an index in this way implies that SQLite has to go through all entries, instead of lookup up the few rows that match some other condition.
When you omit the INDEXED BY clause from your second query, you get the following EXPLAIN QUERY PLAN output:
0 0 0 SEARCH TABLE MessageTable USING INTEGER PRIMARY KEY (rowid=?) (~25 rows)
0 0 0 EXECUTE LIST SUBQUERY 1
1 0 0 SCAN TABLE MessageSearchTable VIRTUAL TABLE INDEX 4: (~0 rows)
0 0 0 USE TEMP B-TREE FOR ORDER BY
which is essentially the same as your first query, except that joins and subqueries are handled slightly differently.
With your table structure, it is not really possible to get faster.
You are doing three operations:
looking up rows in MessageSearchTable;
looking up corresponding rows in MessageTable;
sorting rows by a MessageTable value.
As far as indexes are concerned, steps 2 and 3 conflict with each other.
The database has to choose whether to use an index for step 2 (in which case sorting must be done explicitly) or for step 3 (in which case it has to go through all MessageTable entries).
You could try to return fewer records from the FTS search by making the message time a part of the FTS table and searching only for the last few days (and increasing or dropping the time if you don't get enough results).

Explanation on index on a datetime field and included columns

I have a sqlserver table with the usual
intID(primary key),field1,field2,manyotherfields..., datetime TimeOperation
99% of my different kind of queries start with a TimeOperation BETWEEN startTime AND endTime, and then select * (or count(*)) where fieldA=xxx, and join with other smaller tables.
select * because more or less I need all the fields.
I obviusly created an index on TimeOperation ... but performance are not good enough, so I want to add some index key columns or index included columns, but I'm a little bit confused.
I get the difference between the two, but I don't get how much adding a column in each case impacts on speed and on size.
I guess that the biggest improvement would be to create an index including ALL the columns, is it right? (but I can't afford it in terms of space)
And if I often use field1=xxx for example, adding field1 to the index key columns (after TimeOperation) would give better performance right?
Also...just to be sure how an index with included columns works: if I select rows with TimeOperation in a certain range, sql seeks my TimeOperation index for the rows I'm interested in, and it is faster than scanning all the table because in the index the TimeOperation values are in ascending order, is it right? But then I need all the data now I need all the rest of the data fields of those rows...how does sql acts to retrieve the data? I guess it has a sort of bookmark to those rows in the index, right? But it has to hit the table multiple times then... so including all the columns in the index will save the time to hit the table, it it correct?
Thanks!
Mattia
We will need more information on your table examples of your queries to address this fully, but:
DateTime columns should be highly selective by themselves, so an index with TimeOperation as the first column should address the bulk of queries against TimeOperation.
Do not add all columns blindly to an index, or even on included indexes - this will make the index page density worse and be counter productive (you would be duplicating your table in an index).
If all data in your database centres around TimeOperation, you might consider building your clustered index around it.
If you have queries just on field1 = x then you need a separate index just for field1 (assuming that it is suitably selective), i.e. no TimeOperation on the index if its not in the WHERE clause of your query.
Yes, you are right, when SQL locates a record in an index, it needs to do a key (or RID) lookup back into the cluster to retrieve the rest of the columns. If your non clustered index Includes the other columns in your select statement, the lookup can be avoided. But since you are using SELECT(*), covering indexes are unlikely to help .
Edit
Explanation - Selectivity and density are explained in detail here. e.g. iff your queries against TimeOperation return only a small number of rows (rule of thumb is < 5%, but this isn't always), will the index be used, i.e. your query is selective enough for SQL to choose the index on TimeOperation.
The basic starting point would be:
CREATE TABLE [MyTable]
(
intID INT ID identity(1,1) NOT NULL,
field1 NVARCHAR(20),
-- .. More columns, which may be selected, but not filtered
TimeOperation DateTime,
CONSTRAINT PK_MyTable PRIMARY KEY (IntId)
);
And the basic indexes will be
CREATE NONCLUSTERED INDEX IX_MyTable_1 ON [MyTable](TimeOperation);
CREATE NONCLUSTERED INDEX IX_MyTable_2 ON [MyTable](Field1);
Clustering Consideration / Option
If most of your records are inserted in 'serial' ascending TimeOperation order, i.e. intId and TimeOperation will both increase in tandem, then I would leave the clustering on intID (the default) (i.e. table DDL is PRIMARY KEY CLUSTERED (IntId), which is the default anyway).
However, if there is NO correlation between IntId and TimeOperation, and IF most of your queries are of the form SELECT * FROM [MyTable] WHERE TimeOperation between xx and yy then CREATE CLUSTERED INDEX CL_MyTable ON MyTable(TimeOperation) (and changing PK to PRIMARY KEY NONCLUSTERED (IntId)) should improve this query (Rationale: since contiguous times are kept together, fewer pages need to be read, and the bookmark lookup will be avoided). Even better, if values of TimeOperation are guaranteed to be unique, then CREATE UNIQUE CLUSTERED INDEX CL_MyTable ON MyTable(TimeOperation) will improve density as it will avoid the uniqueifier.
Note - for the rest of this answer, I'm assuming that your IntId and TimeOperations ARE strongly correlated and hence the clustering is by IntId.
Covering Indexes
As others have mentioned, your use of SELECT (*) is bad practice and inter alia means covering indexes won't be of any use (the exception being COUNT(*)).
If your queries weren't SELECT(*), but instead e.g.
SELECT TimeOperation, field1
FROM
WHERE TimeOperation BETWEEN x and y -- and returns < 5% data.
Then altering your index on TimeOperation to include field1
CREATE NONCLUSTERED INDEX IX_MyTable ON [MyTable](TimeOperation) INCLUDE(Field1);
OR adding both to the index (with the most common filter first, or the most selective first if both filters are always present)
CREATE NONCLUSTERED INDEX IX_MyTable ON [MyTable](TimeOperation, Field1);
Either will avoid the rid / key lookup. The second (,) option will address your query where BOTH TimeOperation and Field1 are filtered in a WHERE or HAVING clause.
Re : What's the difference between index on (TimeOperation, Field1) and separate indexes?
e.g.
CREATE NONCLUSTERED INDEX IX_MyTable ON [MyTable](TimeOperation, Field1);
will not be useful for the query
SELECT ... FROM MyTable WHERE Field1 = 'xyz';
The index will only be useful for the queries which have TimeOperation
SELECT ... FROM MyTable WHERE TimeOperation between x and y;
OR
SELECT ... FROM MyTable WHERE TimeOperation between x and y AND Field1 = 'xyz';
Hope this helps?
An index, at its most basic, creates a layer of the "hypertree" structure behind the scenes, which allows the SQL engine to more easily find rows with particular values for indexed columns. Each index creates a different way to "drill down" into the table's data using a binary search (logN performance). Each index you add makes selecting by that index faster, at the cost of slowing insertions/updates (the data must be put in and then indexes must be created).
An index, therefore, should normally be created for combinations of columns that are commonly used to filter records. I would indeed create an index on TimeOperation, and TimeOperation alone.
NEVER simply create an index including all columns of a table, especially a wide one such as this.

Resources