I'm a real beginner when it comes to SQL and I'm currently trying to build a database using postgres. I have a lot of data I want to put into my database in JSON files, but I have trouble converting it into tables. The JSON is nested and contains many variables, but the behavior of jsonb_populate_record allows me to ignore the structure I don't want to deal with right now. So far I have:
CREATE TABLE raw (records JSONB);
COPY raw from 'home/myuser/mydocuments/mydata/data.txt'
create type jsonb_type as (time text, id numeric);
create table test as (
select jsonb_populate_record(null::jsonb_type, raw.records) from raw;
When running the select statement only (without the create table) the data looks great in the GUI I use (DBeaver). However it does not seem to be an actual table as I cannot run select statements like
select time from test;
or similar. The column in my table 'test' also is called 'jsonb_populate_record(jsonb_type)' in the GUI, so something seems to be going wrong there. I do not know how to fix it, I've read about people using lateral joins when using json_populate_record, but due to my limited SQL knowledge I can't understand or replicate what they are doing.
jsonb_populate_record() returns a single column (which is a record).
If you want to get multiple columns, you need to expand the record:
create table test
as
select (jsonb_populate_record(null::jsonb_type, raw.records)).*
from raw;
A "record" is a a data type (that's why you need create type to create one) but one that can contain multiple fields. So if you have a column in a table (or a result) that column in turn contains the fields of that record type. The * then expands the fields in that record.
Related
Is it possible to dynamically select a table by name?
For example I have a table, and every time records are uploaded to it a backup is created first with the date appended to the table name.
table_20191108
table_20191109
table_20191110
table_20191111
What I would like to do is basically write some type of dynamic sql that always
select * from table_MAXDATE
I would like to do this so I can compare table to the most recent backup (e.g. table_20191111) in order to see what changed between the two tables.
haven't tried anything specific yet.
In output I want to select all columns except two columns from a table in q/kdb historical database.
I tried running below query but it does not work on hdb.
delete colid,coltime from table where date=.z.d-1
but it is failing with below error
ERROR: 'par
(trying to update a physically partitioned table)
I referred https://code.kx.com/wiki/Cookbook/ProgrammingIdioms#How_do_I_select_all_the_columns_of_a_table_except_one.3F but no help.
How can we display all columns except for two in kdb historical database?
The reason you are getting par error is due to the fact that it is a partitioned table.
The error is documented here
trying to update a partitioned table
You cannot directly update, delete anything on a partitioned table ( there is a separate db maintenance script for that)
The query you have used as fix is basically selecting the data first in-memory (temporarily) and then deleting the columns, hence it is working.
delete colid,coltime from select from table where date=.z.d-1
You can try the following functional form :
c:cols[t] except `p
?[t;enlist(=;`date;2015.01.01) ;0b;c!c]
Could try a functional select:
?[table;enlist(=;`date;.z.d);0b;{x!x}cols[table]except`colid`coltime]
Here the last argument is a dictionary of column name to column title, which tells the query what to extract. Instead of deleting the columns you specified this selects all but those two, which is the same query more or less.
To see what the functional form of a query is you can run something like:
parse"select colid,coltime from table where date=.z.d"
And it will output the arguments to the functional select.
You can read more on functional selects at code.kx.com.
Only select queries work on partitioned tables, which you resolved by structuring your query where you first selected the table into memory, then deleted the columns you did not want.
If you have a large number of columns and don't want to create a bulky select query you could use a functional select.
?[table;();0b;{x!x}((cols table) except `colid`coltime)]
And show all columns except a subset of columns. The column clause expects a dictionary hence I am using the function {x!x} to convert my list to a dictionary. See more information here
https://code.kx.com/q/ref/funsql/
As nyi mentioned, if you want to permanently delete columns from an historical database you can use the deleteCol function in the dbmaint tools https://github.com/KxSystems/kdb/blob/master/utils/dbmaint.md
I'm working on an application that imports data from Access to SQL Server 2008. Currently, I'm using a stored procedure to import the data individually by record. I can't go with a bulk insert or anything like that because the data is inserted into two related tables...I have a bunch of fields that go into the Account table (first name, last name, etc.) and three fields that will each have a record in an Insurance table, linked back to the Account table by the auto-incrementing AccountID that's selected with SCOPE_IDENTITY in the stored procedure.
Performance isn't very good due to the number of round trips to the database from the application. For this and some other reasons I'm planning to instead use a staging table and import the data from there. Reading up on my options for approaching this, a cursor that executes the same insert stored procedure on the data in the staging table would make sense. However it appears that cursors are evil incarnate and should be avoided.
Is there any way to insert data into one table, retrieve the auto-generated IDs, then insert data for the same records into another table using the corresponding ID, in a set-based operation? Or is a cursor my only option here?
Look at the OUTPUT clause. You should be able to add it to your INSERT statement to do what you want.
BTW, if you need to output columns into the second table that weren't inserted into the first one, then use MERGE instead of INSERT (as suggested in the comment to the original question) as its OUTPUT clause supports referencing other columns from the source table(s). Otherwise, keeping it with an INSERT is more straightforward, and it does give you access to the inserted identity column.
I'm having experiment to worked out in inserting multiple record into related table using databinding. So, try this!
Hopefully this is very helpful. Follow this link How to insert record into related tables. for more information.
I need to use the same Stored Procedures against many tables all with the same structure in my DB. This is data loaded from customers,with one table/customer and the data needs calculations/checks run before it's loaded to our DataWarehouse.
So far these are the options and issues I've found and I'm looking for a better pattern/approach.
Create a view that points to the
table I want to process, the SPs
then talk to that view. This works
well (especially once I'd worked out
how to create views 'automagically'
based on their columns). But the
view can only be used with one table
at a time, forcing the system to
deal with one customer at a time.
Use dynamic sql within each SP -
makes the SPs much harder to
read/debug and for those reasons has
been ruled out
Create a partitioned view across
all the tables and then use a
paramatised table function to return
just the data we're interested in -
ah but then I can't update the data
as the function returns a table that
can be only used for select
Use dynamic sql inside a function
(can't be done) to create a view
(which also can't be done) .... give
up
Within the SP create a temp table
with over the target table using
dynamic sql, but then the temp table
only exists in the session that runs
the dynamic sql not the 'parent'
session that's running the SP ...
give up
Create a global temp table using
dynamic SQL to avoid the scope issue
of 5, then run the SP against the
global temp table. Still run into
the single customer issue.
Create the view as in 1 within a
transaction and then run all the SPs
and then commit - works fine for one
user, but any others are now blocked
trying to create a new view of the
same name
Use a temporary view ... can't in
T/Sql
Move all the code into .Net - but
we have environment issues where
tsql is much easier to host/run
I know I'm not the only person who has this problem, have any of you good people solved it, please help.
Maybe your approach is wrong, I will go deep in details in a while but it seems that your problem can be solved using SSIS
-- Updated answer:
First, the big picture:
The most affordable way to process the tables dynamically is using a script instead of a stored procedure. If you want to make table access randomly chosen, you certainly will not use any of the performance advantages of stored procedures, i.e. execution plans. A SQL Script can be easily upgraded to point one table at runtime using placeholders and replacing it before executing.
The script can be loaded from the filesystem, a variable, a text column in a table, etc. The loading process consists in read the script content to a string variable. This step occurs once.
The next step is the preparation stage. This step will be executed for each table to be processed. The main business of this step is to replace the table placeholders with the current table being processed. Also is possible to set parameter values like any parameter you can need to pass into the sp that you already wrote.
The last step is the execution of the script. As is already loaded into a variable and the placeholders were set to the current table name, you can safely call a ExecuteSQLTask with the sql variable as the input. This process of course happens for each table you want to process.
Ok. Now let's see this in action.
This is a sample database model:
CREATE TABLE [dbo].[t_n](
[id] [int] IDENTITY(1,1) NOT NULL,
[name] [varchar](50) NOT NULL,
[start] [datetime] NULL,
CONSTRAINT [PK_t_n] PRIMARY KEY CLUSTERED ([id] ASC)
) ON [PRIMARY]
where t_n represents any table (t_1, t_2, t_3, etc).
This is your current stored procedure:
CREATE PROCEDURE SpProcessT_n
AS
BEGIN
SET NOCOUNT ON;
SELECT * FROM [t1];
END
GO
Now, transform this stored procedure to a Sql script, placing a placeholder instead of the table name
SET NOCOUNT ON;
SELECT * FROM [$table_name];
I choose to save this in a .sql file in the filesystem to keep the POC as simple as possible.
Next, create a SSIS Package like this:
These are the settings I choose to set up the loop:
And this is the way you can assign the table name to a variable called appropriately _table_name_
This is the setup of the script task, here you find that the variable _table_name_ has read only access, while a new variable called SqlExec has read/write access:
And this is it's Main function:
public void Main()
{
String Table_Name = Dts.Variables["table_name"].Value.ToString();
String SqlScript;
Regex reg = new Regex(#"\$table_name", RegexOptions.Compiled);
using (var f = File.OpenText(#"c:\sqlscript.sql")) {
SqlScript = f.ReadToEnd();
f.Close();
}
SqlScript = reg.Replace(SqlScript, Table_Name);
Dts.Variables["SqlExec"].Value = SqlScript;
Dts.TaskResult = (int)ScriptResults.Success;
}
You can notice that the Dts Variable SqlExec contains the sql script that will be executed. Now you can set the following options in your ExecuteSqlTask:
Successfully tested in MSSQL 2008, if you put a insert inside the script file you will notice new rows in each table.
Hope this helps!
If your application can afford to have one cut-off day late, then you can have a nightly scheduled job to run an SSIS package that will consolidate all 150+ tables into one single huge table. Since the freshness of the results of the queries against that huge table will then be 1 'date' late, this solution will not include any rows that recently been loaded.
You can actually time the running of this package. If it is still amazingly fast, say within 30 minutes, then you can bet to run it in every few hours, like during: the start of work day, lunch break, and end of day. This way you can have a nearly fresh data to query with.
Write a partitioned view including table names?
SELECT 'TableName', t.* FROM TableName t
UNION ALL
SELECT 'TableName2', t.* FROM TableName2 t
Then write a single instead of trigger which uses dynamic SQL for writing (less testing involved with that use of dynamic SQL because you'd just write the simple CRUD operations once for all tables I'd think)
I would not do this with SQL. What you are describing sounds like a traditional ETL situation.
Since all of the customer tables are the same, I would create a table in the data warehouse with all the columns from the client table, a surrogate key column, and a type identifier. You have an option to create a "staging" table here that will only have data in it during the ETL process, or just working on a single "live" table. I would create the staging table.
Then within SSIS package (don't worry you can still schedule from SQL Server agent, it hasn't totally left the DB server), start the ETL process...
E(xtract): copy the data from your source into the staging table in the data warehouse. You most likely want to use a sub-package within a foreach loop and changing the name of the table that you want to process from an external store (most people would say put this in the warehouse, but its up to you).
T(ransform): run the calculations/checks you were talking about, but do it on the whole set...
L(oad): Copy it to your real within the data warehouse.
There are a couple things I would NOT do.
1. Modify the data in the source table.
2. Try to do this in t-sql. Its just not what tsql is good at.
If you need more detail on this approach, I would probably ask the question with some Business Intelligence tags. I'll be traveling for the next week or so, but I will try to look at the comments to clear anything up if you need me to.
I am fairly certain that the standard way to solve this is using dynamic SQL in each sp (your option 2), which has already been ruled out.
Your goal is to make generic, multi-table SQL. I don't see how you intend to accomplish that without sacrificing some efficiency and readability.
Imagine a table with the following structure on PostgreSQL 9.0:
create table raw_fact_table (text varchar(1000));
For the sake of simplification I only mention one text column, in reality it has a dozen. This table has 10 billion rows and each column has lots of duplicates. The table is created from a flat file (csv) using COPY FROM.
To increase performance I want to convert to the following star schema structure:
create table dimension_table (id int, text varchar(1000));
The fact table would then be replaced with a fact table like the following:
create table fact_table (dimension_table_id int);
My current method is to essentially run the following query to create the dimension table:
Create table dimension_table (id int, text varchar(1000), primary key(id));
then to create fill the dimension table I use:
insert into dimension_table (select null, text from raw_fact_table group by text);
Afterwards I need to run the following query:
select id into fact_table from dimension inner join raw_fact_table on (dimension.text = raw_fact_table.text);
Just imagine the horrible performance I get by comparing all strings to all other strings several times.
On MySQL I could run a stored procedure during the COPY FROM. This could create a hash of a string and all subsequent string comparison is done on the hash instead of the long raw string. This does not seem to be possible on PostgreSQL, what do I do then?
Sample data would be a CSV file containing something like this (I use quotes also around integers and doubles):
"lots and lots of text";"3";"1";"2.4";"lots of text";"blabla"
"sometext";"30";"10";"1.0";"lots of text";"blabla"
"somemoretext";"30";"10";"1.0";"lots of text";"fooooooo"
Just imagine the horrible performance
I get by comparing all strings to all
other strings several times.
When you've been doing this a while, you stop imagining performance, and you start measuring it. "Premature optimization is the root of all evil."
What does "billion" mean to you? To me, in the USA, it means 1,000,000,000 (or 1e9). If that's also true for you, you're probably looking at between 1 and 7 terabytes of data.
My current method is to essentially
run the following query to create the
dimension table:
Create table dimension_table (id int, text varchar(1000), primary key(id));
How are you gonna fit 10 billion rows into a table that uses an integer for a primary key? Let's even say that half the rows are duplicates. How does that arithmetic work when you do it?
Don't imagine. Read first. Then test.
Read Data Warehousing with PostgreSQL. I suspect these presentation slides will give you some ideas.
Also read Populating a Database, and consider which suggestions to implement.
Test with a million (1e6) rows, following a "divide and conquer" process. That is, don't try to load a million at a time; write a procedure that breaks it up into smaller chunks. Run
EXPLAIN <sql statement>
You've said you estimate at least 99% duplicate rows. Broadly speaking, there are two ways to get rid of the dupes
Inside a database, not necessarily the same platform you use for production.
Outside a database, in the filesystem, not necessarily the same filesystem you use for production.
If you still have the text files that you loaded, I'd consider first trying outside the database. This awk one-liner will output unique lines from each file. It's relatively economical, in that it makes only one pass over the data.
awk '!arr[$0]++' file_with_dupes > file_without_dupes
If you really have 99% dupes, by the end of this process you should have reduced your 1 to 7 terabytes down to about 50 gigs. And, having done that, you can also number each unique line and create a tab-delimited file before copying it into the data warehouse. That's another one-liner:
awk '{printf("%d\t%s\n", NR, $0);}' file_without_dupes > tab_delimited_file
If you have to do this under Windows, I'd use Cygwin.
If you have to do this in a database, I'd try to avoid using your production database or your production server. But maybe I'm being too cautious. Moving several terabytes around is an expensive thing to do.
But I'd test
SELECT DISTINCT ...
before using GROUP BY. I might be able to do some tests on a large data set for you, but probably not this week. (I don't usually work with terabyte-sized files. It's kind of interesting. If you can wait.)
Just to questions:
- it neccessary to convert your data in 1 or 2 steps?
- May we modify the table while converting?
Running more simplier queries may improve your performance (and the server load while doing it)
One approach would be:
generate dimension_table (If i understand it correctly, you don't have performance problems with this) (maybe with an additional temporary boolean field...)
repeat: choose one previously not selected entry from dimension_table, select every rows from raw_fact_table containing it and insert them into fact_table. Mark dimension_table record as done, and next... You can write this as a stored procedure, and it can convert your data in the background, eating minimal resources...
Or another (probably better):
create fact_table as EVERY record from raw_fact_table AND one dimension_id. (so including dimension_text and dimension_id rows)
create dimension_table
create an after insert trigger for fact_table which:
searches for dimension_text in fact_table
if not found, creates a new record in dimension_table
updates dimension_id to this id
in a simle loop, insert every record from raw_fact_table to fact_table
You are omitting some details there at the end, but I don't see that there necessarily is a problem. It is not in evidence that all strings are actually compared to all other strings. If you do a join, PostgreSQL could very well pick a smarter join algorithm, such as a hash join, which might give you the same hashing that you are implementing yourself in your MySQL solution. (Again, your details are hazy on that.)
-- add unique index
CREATE UNIQUE INDEX uidx ON dimension_table USING hash(text);
-- for non case-sensitive hash(upper(text))
try hash(text); and btree(text) to see which one is faster
I an see several ways of solving your problem
There is md5 function in PostgreSql
md5(string) Calculates the MD5 hash of string, returning the result in hexadecimal
insert into dimension_table (select null, md5(text), text from raw_fact_table group by text)
add md5 field into raw_fact_table as well
select id into fact_table from dimension inner join raw_fact_table on (dimension.md5 = raw_fact_table.md5);
Indexes on MD5 filed might help as well
Or you can calculate MD5 on the fly while loading the data.
For example our ETL tool Advanced ETL processor can do it for you.
Plus it can load data into multiple tables same time.
There is a number of on-line tutorials available on our web site
For example this one demonstrates loading slow changing dimension
http://www.dbsoftlab.com/online-tutorials/advanced-etl-processor/advanced-etl-processor-working-with-slow-changing-dimension-part-2.html