SQL Tips¶

General HA¶

Timestamps: Home Assistant stores times as Unix timestamps (floating point). Use functions to convert to readable dates.
Entity relationships: Always join states with states_meta to get the actual entity_id, and optionally with state_attributes for attributes.
Performance: For large databases, add WHERE clauses to limit time ranges, especially when querying states or events.
JSON in attributes: The shared_attrs field contains JSON data. You can parse it with json_extract() function.

SQLite MySQL Dialect Differences¶

	SQLite	MariaDB
Timestamp	`datetime(timestamp, 'unixepoch')`	`FROM_UNIXTIME(timestamp)`
String Concatenation	`'Hello' ??`	`CONCAT('Hello', ' ', 'World')`
Limit/Pagination	`LIMIT 10 OFFSET 20`	`LIMIT 10 OFFSET 20`
Auto-increment Primary Key	`INTEGER PRIMARY KEY` (automatic)	`AUTO_INCREMENT`
Boolean Type	No native BOOLEAN, uses INTEGER (0/1)	`BOOLEAN` (stored as TINYINT)
Case-Insensitive Search	`LIKE` (case-insensitive by default)	`LIKE` (case-insensitive) or `LIKE BINARY`
Now	SELECT datetime('now')	SELECT NOW()
Adding days to a date	SELECT datetime('now', '+7 days')	SELECT DATE_ADD(NOW(), INTERVAL 7 DAY)
String length	SELECT length(column_name)	SELECT LENGTH(column_name) -- Same, but case matters in some contexts

Conditional logic (CASE) - mostly the same¶

-- Both SQLite and MariaDB (ANSI standard)
SELECT 
    CASE 
        WHEN value > 10 THEN 'high'
        ELSE 'low'
    END as category

Grouping with aggregation¶

-- Both work the same for basic aggregates
SELECT category, COUNT(*), AVG(value)
FROM table
GROUP BY category

Portability Tips¶

Most Portable (Works Everywhere)¶

SELECT column1, column2
FROM table1
JOIN table2 ON table1.id = table2.id
WHERE column1 = 'value'
GROUP BY column1
ORDER BY column2
LIMIT 10

Less Portable (Database-Specific)¶

Date/time functions
String manipulation
Regular expressions
Window functions (support varies)
JSON operations
Full-text search

JSON¶

	SQLite	MariaDB
Get specific field	json_extract()	JSON_EXTRACT()
Iterate top level keys: returns table with key/value/type	json_each()	json_each()
Iterate all nested paths: returns table with fullkey/path/value	json_tree()	json_tree()
Get keys	Use json_each()	json_keys()
Format JSON nicely for display purposes	json_pretty()	json_pretty()
Get value type: 'text', 'integer', 'real', etc.	json_type()	json_type()

Filter by JSON Values (MySQL)¶

SELECT 
    sm.entity_id,
    s.state,
    FROM_UNIXTIME(s.last_updated_ts) as last_updated,
    sa.shared_attrs->>'$.latitude' as latitude,
    sa.shared_attrs->>'$.longitude' as longitude
FROM states s
INNER JOIN states_meta sm ON s.metadata_id = sm.metadata_id
LEFT JOIN state_attributes sa ON s.attributes_id = sa.attributes_id
WHERE sm.entity_id = 'device_tracker.sm_p620'
  AND JSON_EXTRACT(sa.shared_attrs, '$.latitude') IS NOT NULL
  AND s.state = 'not_home'
ORDER BY s.last_updated_ts DESC;

Cast to Numeric Type (MySQL)¶

If you need to do math with the extracted values:

SELECT 
    sm.entity_id,
    FROM_UNIXTIME(s.last_updated_ts) as last_updated,
    CAST(sa.shared_attrs->>'$.latitude' AS DECIMAL(10, 6)) as latitude,
    CAST(sa.shared_attrs->>'$.longitude' AS DECIMAL(10, 6)) as longitude,
    CAST(sa.shared_attrs->>'$.gps_accuracy' AS UNSIGNED) as gps_accuracy
FROM states s
INNER JOIN states_meta sm ON s.metadata_id = sm.metadata_id
LEFT JOIN state_attributes sa ON s.attributes_id = sa.attributes_id
WHERE sm.entity_id = 'device_tracker.sm_p620'
ORDER BY s.last_updated_ts DESC
LIMIT 20;

Pattern Matching¶

Find all entities containing "temperature"¶

SELECT DISTINCT sm.entity_id
FROM states_meta sm
WHERE sm.entity_id LIKE '%temperature%'
ORDER BY sm.entity_id;

Common pattern matching wildcards¶

% = matches any sequence of characters (including zero characters)
_ = matches exactly one character

Pattern Matching Examples¶

Entities that START with "sensor.temperature"¶

SELECT DISTINCT sm.entity_id
FROM states_meta sm
WHERE sm.entity_id LIKE 'sensor.temperature%'
ORDER BY sm.entity_id;

Entities that END with "temperature"¶

SELECT DISTINCT sm.entity_id
FROM states_meta sm
WHERE sm.entity_id LIKE '%temperature'
ORDER BY sm.entity_id;

Entities containing "temp" OR "temperature"¶

SELECT DISTINCT sm.entity_id
FROM states_meta sm
WHERE sm.entity_id LIKE '%temp%' 
   OR sm.entity_id LIKE '%temperature%'
ORDER BY sm.entity_id;

Case-insensitive matching (SQLite default behavior)¶

-- These will match "Temperature", "TEMPERATURE", "temperature", etc.
SELECT DISTINCT sm.entity_id
FROM states_meta sm
WHERE sm.entity_id LIKE '%temperature%'sql
ORDER BY sm.entity_id;

Match exact pattern with underscore (single character wildcard)¶

-- Matches "sensor.temp_1", "sensor.temp_2", etc.
SELECT DISTINCT sm.entity_id
FROM states_meta sm
WHERE sm.entity_id LIKE 'sensor.temp_%'
ORDER BY sm.entity_id;

Usage of SQL View¶

A view in SQL is like a saved query that acts as a virtual table. It doesn't store data itself—it just stores the query definition and runs it whenever you access the view.

Think of It Like a Shortcut¶

Instead of writing a complex query every time, you save it as a view and then query the view like it's a table.

Without a View (Long Query Every Time)¶

-- You have to write this complex query every time:
SELECT 
    sm.entity_id,
    s.state,
    datetime(s.last_updated_ts, 'unixepoch', 'localtime') as last_updated,
    json_extract(sa.shared_attrs, '$.latitude') as latitude,
    json_extract(sa.shared_attrs, '$.longitude') as longitude
FROM states s
INNER JOIN states_meta sm ON s.metadata_id = sm.metadata_id
LEFT JOIN state_attributes sa ON s.attributes_id = sa.attributes_id
WHERE sm.entity_id = 'device_tracker.sm_p620'
ORDER BY s.last_updated_ts DESC;

With a View (Simple Query)¶

-- Create the view once:
CREATE VIEW device_locations AS
SELECT 
    sm.entity_id,
    s.state,
    datetime(s.last_updated_ts, 'unixepoch', 'localtime') as last_updated,
    json_extract(sa.shared_attrs, '$.latitude') as latitude,
    json_extract(sa.shared_attrs, '$.longitude') as longitude
FROM states s
INNER JOIN states_meta sm ON s.metadata_id = sm.metadata_id
LEFT JOIN state_attributes sa ON s.attributes_id = sa.attributes_id
WHERE sm.entity_id LIKE 'device_tracker.%';

-- Then query it like a regular table:
SELECT * FROM device_locations 
WHERE entity_id = 'device_tracker.sm_p620'
ORDER BY last_updated DESC
LIMIT 10;

Key Characteristics of Views¶

Aspect	Description
Virtual table	Looks like a table, but doesn't store data
Always fresh	Shows current data from underlying tables
Simplifies queries	Hide complex JOINs and calculations
Reusable	Write once, use many times
No extra storage	Only stores the query definition
Performance	Same as running the underlying query

Real-World Analogy¶

View = Saved search/bookmark

A browser bookmark doesn't store the webpage
It just stores the URL (the "query")
When you click it, it fetches fresh content
Same with views: they execute the query and show fresh data

When to Use Views¶

✅ Good Use Cases¶

Simplify complex queries you use often

CREATE VIEW latest_temperatures AS
SELECT 
    sm.entity_id,
    s.state as temperature,
    datetime(s.last_updated_ts, 'unixepoch', 'localtime') as last_updated
FROM states s
INNER JOIN states_meta sm ON s.metadata_id = sm.metadata_id
WHERE sm.entity_id LIKE 'sensor.%temperature%'
  AND s.state NOT IN ('unavailable', 'unknown');

Security/permissions - Show only certain columns to users

CREATE VIEW public_device_status AS
SELECT entity_id, state, last_updated_ts
FROM states
-- Excludes sensitive attributes

Hide implementation details

CREATE VIEW energy_consumption AS
SELECT 
    statistic_id,
    sum as total_kwh,
    datetime(start_ts, 'unixepoch') as date
FROM statistics s
INNER JOIN statistics_meta sm ON s.metadata_id = sm.id
WHERE sm.unit_of_measurement = 'kWh';

❌ When NOT to Use Views¶

One-time queries - Just run the query directly
Very slow queries - Views don't cache results; they recalculate every time
Temporary exploration - Use CTEs (WITH clause) instead

View Operations¶

Create a View¶

CREATE VIEW my_view_name AS
SELECT ...;

Query a View¶

SELECT * FROM my_view_name WHERE ...;

See All Views¶

SELECT name FROM sqlite_master WHERE type = 'view';

See View Definition¶

SELECT sql FROM sqlite_master WHERE type = 'view' AND name = 'my_view_name';

Drop a View¶

DROP VIEW my_view_name;

Replace a View¶

DROP VIEW IF EXISTS my_view_name;
CREATE VIEW my_view_name AS
SELECT ...;

Example: Home Assistant Use Case¶

Let's create a view for all entities with statistics:

CREATE VIEW entities_with_statistics AS
SELECT 
    sm.entity_id,
    s.state,
    datetime(s.last_updated_ts, 'unixepoch', 'localtime') as last_updated,
    json_extract(sa.shared_attrs, '$.state_class') as state_class,
    json_extract(sa.shared_attrs, '$.unit_of_measurement') as unit,
    json_extract(sa.shared_attrs, '$.device_class') as device_class
FROM states s
INNER JOIN states_meta sm ON s.metadata_id = sm.metadata_id
INNER JOIN (
    SELECT metadata_id, MAX(state_id) as max_state_id
    FROM states
    GROUP BY metadata_id
) latest ON s.metadata_id = latest.metadata_id AND s.state_id = latest.max_state_id
LEFT JOIN state_attributes sa ON s.attributes_id = sa.attributes_id
WHERE json_extract(sa.shared_attrs, '$.state_class') IS NOT NULL;

Now you can easily query it:

-- Get all energy sensors
SELECT * FROM entities_with_statistics 
WHERE device_class = 'energy';

-- Get all measurement sensors
SELECT * FROM entities_with_statistics 
WHERE state_class = 'measurement';

-- Get all sensors with no unit
SELECT * FROM entities_with_statistics 
WHERE unit IS NULL;

Summary¶

View = Saved query that acts like a virtual table

✅ Simplifies complex queries
✅ Always shows fresh data
✅ No extra storage needed
✅ Makes repeated queries easier
❌ Doesn't cache/store results
❌ Can be slower if underlying query is complex

For Home Assistant database exploration, views are great for queries you run repeatedly (like "show me all statistics entities" or "show me all device tracker locations").