Develop Reference

Unnest (or similar) in SQLite

I'd like to return multiple rows as a single row to be handled by an Android Cursor Adapter.
I currently have a table with a date column and I'd like to return, as a single row, all the rows that have the same date.
Consider the following table:
ID | Name | Date
-------------------------------
1 | 'Mark' | '08/06/15'
2 | 'Peter' | '08/06/15'
3 | 'Henry' | '08/06/15'
4 | 'Bob' | '17/04/16'
5 | 'Tony' | '23/08/13'
6 | 'Tim' | '17/04/16'
I'd like to query the results as follows:
Date | Names
------------------------------------
'08/06/15' | 'Mark, Peter, Henry'
'17/04/16' | 'Bob, Tim'
'23/08/13' | 'Tony'
Using this link I was able to obtain the following query:
SELECT t1.id, GROUP_CONCAT(t1.Name ) AS Names
FROM Table1 t1 JOIN Table2 t2 ON t1.ID = t2.ID
GROUP BY t1.ID;
However, since all the data is from the same table, and I know the dates to query in advance, I was hoping to use JOIN with an inputted array and have SQLite parse it (ie. like unnest in Postgres) as if it came from an actual table.
The array would be something like:
['08/06/15', '17/04/16', '23/08/13', '09/08/18']
This can probably also be done by nesting SQL queries, but I'd like an optimized solution if possible.

SQLite has no array type.
The simplest way would be to use the IN operator:
SELECT ID, group_concat(Name) AS Names
FROM MyTable
WHERE Date IN ('08/06/15', '17/04/16', '23/08/13', '09/08/18')
GROUP BY ID;
If you really want to use a join, you can construct a virtual table with a VALUES clause (in Android Jelly Bean or later), or a compound query:
SELECT ID, group_concat(Name) AS Names
FROM MyTable
JOIN (VALUES('08/06/15'), ('17/04/16'), ('23/08/13'), ('09/08/18'))
ON Date = column1
GROUP BY ID;
SELECT ID, group_concat(Name) AS Names
FROM MyTable
JOIN (SELECT '08/06/15' AS Date UNION ALL
SELECT '17/04/16' UNION ALL
SELECT '23/08/13' UNION ALL
SELECT '09/08/18' ) USING (Date)
GROUP BY ID;

Is this what you want:
SELECT Date, group_concat(Name) AS Names
FROM table_name
GROUP BY Date;

Sqlite Distinct Query From -To, To - From

I have following data:
me = my_userid;
Table: Message
message_id | message_from | message_to
1 | me | user
2 | user | me
Running query gives two rows (SELECT DISTINCT message_from,message_to FROM Message WHERE message_from ='"+me+"' OR message_to ='"+me+"'");
Used OR because my id can be in from(when sending) or in TO (when other user sends me message)
-- This returns two rows however i want it to return one row because if you switch to - from you get same ids, So how can this be done in the query. Thanks

You can use min() and max() as scalar functions:
select distinct min(message_from, message_to) as m1, max(message_from, message_to) as m2
from message
where message_from ='"+me+"' OR message_to ='"+me+"'"
These are equivalent to least() and greatest() in other databases.

Group by in Sqlite?

I have 2 tables Products and Invoices. Each Products can be in some Invoices.
Product:
Id| Name
ـــــــــــــــ
1 | pencil
2 | pen
3 | ruler
Invoice:
Id| ProductID| Serial |Qty
ـــــــــــــــــــــــــــــ
1 | 1 |100 |5
2 | 1 |200 |6
3 | 2 |300 |8
4 | 3 |400 |18
When I write the following query
select * , SUM(Invoices.Qty)
FROM Products left outer join Invoices on Products.Id = Invoices.ProductID
Group by Products.Id
result:
Id| Name |Id|ProductID| Serial|SUM(Invoices.Qty)
ـــــــــــــــــــــــــــــــــــــــــــــــــــــ
1 |pencil |2 |1 |200 |11 -> (5+6)
2 |pen |3 |2 |300 |8
3 |ruler |4 |3 |400 |18
it returns a list of all products that is joined by the last Invoice. It calculates the Qty correctly but the problem is the selection of the last Invoice. How can I write a query that calculates Qty correctly and returns the Invoices with the id that I want.
in this example I want the invoice id 1 (serial 100) not 2 (serial 200)

Try this:
SELECT *,
(select sum(i3.qty) from Invoices i3 where i3.productId = i.productId)
FROM invoices I
JOIN products P
on I.productId = P.id
WHERE NOT EXISTS(
SELECT 'PREVIOUS'
FROM invoices I2
where I2.ProductId = I.productId
and I2.id < I.id
)
In this way you get the first invoice about a product. I don't understand if doesn't exist an invoice in you want to show the product. If yes, you must change the query with a left outer join

In SQLite 3.7.11 or later, you can use the MIN or MAX functions to force a specific from the group to be returned:
SELECT *, MIN(Invoices.Serial), SUM(Invoices.Qty)
FROM Products LEFT JOIN Invoices ON Products.Id = Invoices.ProductID
GROUP BY Products.Id

Return rows around my selection args ("surrounding rows") in a database query

I'm querying a database in Android. The table is your ordinary table with values, nothing special.
What I need: return the two events that happened before and after the given timestamp.
Example: let's suppose I have the table below, and my selection is 1332200003002:
_id | Time | Value
... ...
n | 1332200002000 | 145
n+1 | 1332200003001 | 98 (this is <= selection)
(1332200003002 is here, between those two)
n+2 | 1332200004000 | 90 (this is > selection)
n+3 | 1332200005000 | 100
n+4 | 1332200005001 | 280
So, if my selection is 1332200003001, or 1332200003002... I'd want the returned rows to be n+1 and n+2, so that I can see that the Value went from 98 to 90.
What I'm using is a CursorLoader, so it must preferably fit into its usual call.
My code size thanks you!
As a side note, I can guess safe values for BETWEEN (it IS working already), and then iterate the few remaining Cursor rows in Java to pinpoint the two rows that I need. However, this seems to me like a very common need, hence the question. Seems a waste to do it in Java with all those usual bumper tests we need to do with a Cursor.

SELECT *
FROM myTable LIMIT 2
OFFSET
(SELECT _id
FROM myTable
WHERE time<=1332200003002
ORDER BY time DESC LIMIT 1) - 1;
What this does:
Select 2 entries from the table. The offset of the first entry is selected as follows:
Choose the latest row where time <= 1332200003002, and calculate its offset from the first row.
The -1 at the end is needed if your _id values start at 1 rather than 0. (Change this value as needed to convert your _id values into zero-based offsets.)

How to run DISTINCT on non-key in SQL

I have a database that can have similar rows, but have different keys and a different boolean column. Here is what the database looks like:
columns: _id, name, img, address, available
Two entries can look like this:
_id | name | img | address | available
-------------------------------------------------------
1 | John | http://img.com/222 | 1 Main St | 1
2 | John | http://img.com/222 | 1 Main St | 0
I want a query that will give me all results that have a distinct key, and if there are duplicate entries(ignoring the fact that _id would be different), it will give back only the first one. Here is the query I have:
SELECT p1.*
FROM (SELECT DISTINCT _id, available FROM people) p
INNER JOIN people p1
ON p1._id=p._id
ORDER BY p1.available DESC;
I know this isn't right, but maybe it explains a little what I am looking for. Would I want to use GROUP BY here?

I want a query that will give me all results that have a distinct key, and if there are duplicate entries(ignoring the fact that _id would be different), it will give back only the first one.....the _id isn't what I want to be distinct, as they [the ids] are already unique. ... . Ideally it will order by 'available' in descending order so that if there are two columns with the same data(aside from _id and available), it will return the row with '1' for the available column
select name, image, address, max(availability) as avail
from T
group by name, image, address
Then you can join the set returned by the query above, as an inline view, to your table:
select * from T
inner join
(
select name, image, address, max(availability) avail
from T
group by name, image, address
) as foo
on T.name = foo.name and T.image = foo.image and T.address = foo.address and T.availability = foo.avail
It would help to have a composite index so: (name, image, address).
Caveat: if there is more than one row where a specific {name, image, address} triad has availablility =1, the query will return multiple rows for the triad:
2 | John | http://img.com/222 | 1 Main St | 1
6 | John | http://img.com/222 | 1 Main St | 1
P.S. It sounds as though you wished the triad (name, image, address) had been created in your table an alternate UNIQUE key.

this sql may solve your problem:
select b.* from （select distinct _id from people) a, people b where a._id = b._id order by b.available

I actually just asked a similar question and received a great answer from an experienced user here:
SQL Populating with distinct data and a sequence
Based on what he told me, perhaps this query would provide you with what you want:
SELECT p1.*
FROM (SELECT DISTINCT _id, name from people) p
INNER JOIN people p1
ON p1._id=p._id
ORDER BY p1.available desc
apologies if that's a fail and doesn't work!
EDIT: It just occurred to me that I have no idea which distinct name+_id combo this will extract.. the available=1 or the available=0 or a random selection..! Let me know what happens anyway..

If you want the first row which has the lowest _id among those that have the highest available value (between 1 and 0), you can "record" the _id inside the aggregated value generated by the grouping.
The value to compare is constructed in a way that orders the record by their available field in descending order and then by their _id field in descending order, and allow to easily retrieve the value of the _id with the modulo operator (assuming available max value is 1 and the ids are never above 100000000).
select people.* from people
inner join (
select name, img, address,
min((1-available)*100000000 + _id) avail_id
from people group by name, img, address
) as foo on people._id = foo.avail_id % 100000000;
I adapted it Tim's query.
You can also do that without subquery:
select people.* from people
left outer join people as other on
other.name = people.name and
other.img = people.img and
people.address=other.address and
(1 - people.available) * 100000000 + people._id >
(1 - other.available) * 100000000 + other._id
where other.available is null;