Graph Database with T-SQL

SQL Server 2017 will bring users graph database. Graph database is type of database that uses graph attributes – edges (or relationships) and nodes (or vertices) and features, to store data. Relationship between nodes are through edges and features (also called properties) hold the values to represent the strength or direction of the edges. Many of this links can be retrieved easy with a single operation (in comparison to T-SQL language, which can quickly become complex).

Links are also stored in the database, and when querying the graph database, use join clause, to collect the related data. Especially very useful for hierarchical structures and complex social structures, can be easily stored and queried in comparison to “relational” database.

Network database vs. graph database are similar, whereas, network might have problems when dealing with several chains of edges.

When to use graph/network database should be obvious; when you have hierarchical data or when your data is too complex with many many-to-many relations, that would make a relational database hard to query.

 

To put the graph database to the test, I took bunch of emails from a particular MVP SQL Server distribution list (content will not be shown and all the names will be anonymized). On my gmail account, I have downloaded some 90MiB of emails in mbox file format. With some python scripting,  only FROM and SUBJECTS were extracted:

writer.writerow(['from','subject'])
for index, message in enumerate(mailbox.mbox(infile)):
    content = get_content(message)
    row = [
        message['from'].strip('>').split('<')[-1],
        decode_header(message['subject'])[0][0],"|"
          ]
    writer.writerow(row)

Up to this point, I have extracted a CSV file with only From field and subject field, looking as:

After importing the CSV file to SQL Server database, I have started to prepare my graph database structures:

CREATE TABLE MVP_DataPlatform
(
 MVP_ID INT IDENTITY(1,1) NOT NULL
 ,MVP_Name VARCHAR(100) NULL
 ,MVP_Email VARCHAR(100) NOT NULL
) AS NODE
GO

CREATE TABLE MVP_Topics
(
 Topic_ID INT IDENTITY(1,1) NOT NULL
 ,Title NVARCHAR(250) NOT NULL
) AS NODE
GO

CREATE TABLE Initiated AS EDGE;
CREATE TABLE Replied AS EDGE;

And prepare the data with:

with mails 
as
(
SELECT 
 froms 
 ,REPLACE(SUBSTRING(froms, 1, CASE CHARINDEX(';', froms) 
  WHEN 0 THEN LEN(froms) ELSE CHARINDEX(';',froms)-1 END),',','') AS from_MVP
 ,REPLACE(SUBSTRING(froms, CHARINDEX(';',froms) + 1, 1000),',|','') AS title

FROM [dbo].[clean_mail5_2] 
)
SELECT 
 from_MVP
 ,SUBSTRING(title,2, LEN(title)) as title
 ,CASE WHEN (
 SUBSTRING(title,1, 10) LIKE '%RE%' 
 OR SUBSTRING(title,1, 10) LIKE '%AW%'
 OR SUBSTRING(title,1, 10) LIKE '%FWD%'
 OR SUBSTRING(title,1, 10) LIKE '%OT%'
 ) THEN 0 ELSE 1 END MVP_start
INTO MVP_Mails
FROM mails
 WHERE
 from_MVP like '%@%'

Populating the nodes (vertices) should be a easy task:

INSERT INTO MVP_DataPlatform (MVP_Name, MVP_Email)
SELECT 
 SUBSTRING(from_MVP, 1, CHARINDEX('@',from_MVP)-1) AS MVP_NAME
 ,from_MVP AS MVP_Email
FROM MVP_Mails
GROUP BY 
 SUBSTRING(from_MVP, 1, CHARINDEX('@',from_MVP)-1)
 ,from_MVP;
GO

INSERT INTO MVP_Topics (Title)
SELECT 
 Title
FROM MVP_Mails
GROUP BY 
 title;
GO

But populating the edges (relationships) is part of how you want to define your graph.

I have created two edges; Replied and initiated. When using replied, logic will be following:

1. AB -> send a mail to distribution list =  AB -> AB
2. CD -> replies to AB  = CD -> AB
3. EF -> replies to CD (and AB) = EF -> CD
4. and so on.

When Initiated, logic is slightly different:

1. AB -> send a mail to distribution list =  AB -> AB
2. CD -> replies to AB  = CD -> AB
3. EF -> replies to CD (and AB) = EF -> AB
4. and so on…

So difference is, how I want to store information at step 3.; when EF replies. But this rule is – totally arbitrary and up to your type of business question, hierarchy or many-to-many rules. I can even decide to store information at step 3. normalized, as: ((EF -> CD) AND (EF -> AB)).

To store this set of information into the table of edges, following transformation is needed:

CREATE OR ALTER PROCEDURE INSERT_EDGE
AS
DECLARE @id1 INT;
DECLARE @id2 INT;
DECLARE @i INT;
DECLARE @hid INT;
DECLARE @START INT;

select 
 mail.from_MVP as mail_ID
,pl.MVP_id AS pl_ID
,mail.title
,mail.MVP_Start
,row_number() OVER (partition by title ORDER BY title, MVP_Start DESC) as rn_title
,row_number() Over (order by (select 1)) as rn_global
into #temp
from MVP_Mails as mail
JOIN MVP_DataPlatform AS Pl
ON mail.from_mvp = pl.mvp_email

SELECT @i = MAX(rn_global) FROM #TEMP

SET @hid=1;
WHILE @hid <= @i
BEGIN 
 SELECT @start = MVP_start FROM #temp WHERE rn_global = @hid
 PRINT @start
 SELECT @id1 = pl_id FROM #temp WHERE rn_global = @hid
 PRINT @id1
 IF @start = 1
 SELECT @id2 = pl_id FROM #temp WHERE rn_global = @hid+1
 IF @start = 0
 SET @id2 = @id1
 PRINT @id2

INSERT INTO replied VALUES ((SELECT $NODE_ID FROM MVP_DataPlatform WHERE MVP_ID = @id1), (SELECT $NODE_ID FROM MVP_DataPlatform WHERE MVP_ID = @id2));
 SET @hid = @hid + 1;
END

Executing this procedure:

EXECUTE PROCEDURE INSERT_EDGE

You should be getting a table with relations:

And now you can perform simple queries against the graph – including complex relations, in order to get results relatively simple:

SELECT 
 MVP_1.MVP_Name 
 ,MVP_2.MVP_Name
 ,T.Title
 ,T.Topic_ID
FROM 
 MVP_DataPlatform AS MVP_1
 ,MVP_DataPlatform AS MVP_2
 ,Replied
 ,MVP_Topics AS T
WHERE 
 MATCH(MVP_1-(Replied)->MVP_2)

 

Sure, there are many other ways to do the same steps and also many other ways how you want your data to be represented in a graph.

Happy T-SQL Querying! 🙂

 

Installing R packages with rxInstallPackages in Microsoft R Server

In MicrosoftML package comes – in my opinion – long anticipated function for installing R packages for SQL Server and Microsoft R Server. And, I am super happy.

Last year, in one of my previous blog posts, I have been showing how to install R package from SSMS using sp_execute_external_script. Now, with new package MicrosoftML (that is part of Microsoft R Server 9.x and above)  new function is available that enables you to easy install the package and also little bit more.

Code is relatively simple and straightforward:

USE SQLR;
GO

EXECUTE sp_execute_external_script
 @language = N'R'
 ,@script = N'

 packagesToInstall <- c("caret","tree","party")
 library(MicrosoftML)
 SqlServerCC <- RxInSqlServer(connectionString = "Driver=SQL Server;
+Server=SICN-KASTRUN\\SQLSERVER2017C2;Database=SQLR;
+Trusted_Connection=True;")
 rxInstallPackages(pkgs = packagesToInstall, owner = '', 
+scope = "shared", computeContext = "SqlServerCC");';
GO

This is way too easy to be true, but it is. Make sure to do couple of things prior to running this code:

1. set the compute environment to where your packages are installed
2. set up the correct permissions and access
3. Check up also the tcp/ip protocols

In rxInstallPackages function use computeContext parameter to set either to “Local” or to your  “SqlServer” environment, you can also use scope as shared or private (difference is, if you install package as shared it can be used by different users across different databases, respectively for private). You can also specify owner if you are running this command out of db_owner role.

Happy SQLR-ing!

Performance differences between RevoScaleR, ColumnStore Table and In-Memory OLTP Table

Running *.XDF files using RevoScaleR computational functions versus have dataset available in Columnstore table or in In-Memory OLTP table will be focus of comparison for this blog post.

For this test, I will use the AirLines dataset, available here. Deliberately, I have picked a sample 200 MB (of 13GB dataset) in order to properly test the differences and what should be the best way.

After unzipping the file, I will use following T-SQL query to import the file into SQL Server.

With this example, you can import xdf file directly to SQL Server table (note, that I have transformed a CSV file into XDF and import xdf file into SQL table):

-- must have a write permissions on folder: C:/Program Files/Microsoft SQL Server/130/R_SERVER/library/RevoScaleR/SampleData
DECLARE @RScript nvarchar(max)
SET @RScript = N'library(RevoScaleR)
                rxOptions(sampleDataDir = "C:/Program Files/Microsoft SQL Server/130/R_SERVER/library/RevoScaleR/SampleData")
                inFile <- file.path(rxGetOption("sampleDataDir"), "airsample.csv")
                of <-  rxDataStep(inData = inFile, outFile = "C:/Program Files/Microsoft SQL Server/130/R_SERVER/library/RevoScaleR/SampleData/airline20170428_2.xdf", 
                             transformVars = c("ArrDelay", "CRSDepTime","DayOfWeek")
                            ,transforms = list(ArrDelay = as.integer(ArrDelay), CRSDepTime = as.numeric(CRSDepTime), DayOfWeek = as.character(DayOfWeek))
                            ,overwrite = TRUE
                            ,maxRowsByCols = 10000000
                            ,rowsPerRead = 200000)
                OutputDataSet <- rxXdfToDataFrame(of)'

DECLARE @SQLScript nvarchar(max)
SET @SQLScript = N'SELECT 1 AS N'

EXECUTE sp_execute_external_script
     @language = N'R'
    ,@script = @RScript
    ,@input_data_1 = @SQLScript
WITH RESULT SETS ((ArrDelay INT
                    ,CRSDepTime DECIMAL(6,4)
                    ,DofWeek NVARCHAR(20)))
GO

 

So the whole process is to be done by creating a table, converting the above sp_execute_external_script into procedure and import results from external procedure to the table.

--Complete process
CREATE TABLE AirFlights_small 
(id INT IDENTITY(1,1)
,ArrDelay INT
,CRSDepTime DECIMAL(6,4)
,DofWeek NVARCHAR(20) 
);
GO

CREATE Procedure ImportXDFtoSQLTable
AS
DECLARE @RScript nvarchar(max)
SET @RScript = N'library(RevoScaleR)
                rxOptions(sampleDataDir = "C:/Program Files/Microsoft SQL Server/130/R_SERVER/library/RevoScaleR/SampleData")
                inFile <- file.path(rxGetOption("sampleDataDir"), "airsample.csv")
                of <-  rxDataStep(inData = inFile, outFile = "airline20170428_2.xdf", 
                transformVars = c("ArrDelay", "CRSDepTime","DayOfWeek")
            ,transforms = list(ArrDelay = as.integer(ArrDelay), CRSDepTime = as.numeric(CRSDepTime), DayOfWeek = as.character(DayOfWeek))
            ,overwrite = TRUE
            ,maxRowsByCols = 10000000)
             OutputDataSet <- data.frame(rxReadXdf(file=of, varsToKeep=c("ArrDelay", "CRSDepTime","DayOfWeek")))'
DECLARE @SQLScript nvarchar(max)
SET @SQLScript = N'SELECT 1 AS N'
EXECUTE sp_execute_external_script
     @language = N'R'
    ,@script = @RScript
    ,@input_data_1 = @SQLScript
WITH RESULT SETS ((ArrDelay INT,CRSDepTime DECIMAL(6,4),DofWeek NVARCHAR(20)));
GO

INSERT INTO AirFlights_small
EXECUTE ImportXDFtoSQLTable;
GO

 

There you go. Data are in T-SQL Table. Now we can start with comparisons.  I will be measuring the time to get average air delay time per day of the week.

RevoScaleR

With using the RevoScaleR package, I will be using rxCrossTabs function with the help of transform argument to convert day of the week into factors:

#importing data
outFile2 <- rxDataStep(inData = inFile, outFile = "C:/Program Files/Microsoft SQL Server/130/R_SERVER/library/RevoScaleR/SampleData/airline20170428_2.xdf", 
            transformVars = c("ArrDelay", "CRSDepTime","DayOfWeek")
           ,transforms = list(ArrDelay = as.integer(ArrDelay), CRSDepTime = as.numeric(CRSDepTime), DayOfWeek = as.character(DayOfWeek))
           ,overwrite = TRUE
           ,maxRowsByCols = 10000000)

of2 <- data.frame(rxReadXdf(file=outFile2, varsToKeep=c("ArrDelay", "CRSDepTime","DayOfWeek")))

summary(rxCrossTabs(ArrDelay~DayOfWeek
                    ,data = of2  #outFile2
                    ,transforms = transforms
                    ,blocksPerRead=300000), output="means")

Now get those times:

# Getting times
system.time({ 
  summary(rxCrossTabs(ArrDelay~DayOfWeek
                      ,data = of2
                      ,transforms = transforms
                      ,blocksPerRead=300000), output="means")
  })

With results of 7.8 on elapsed time and computation time of 3.8 second.

Rows Read: 8400013, Total Rows Processed: 8400013, Total Chunk Time: 3.825 seconds 
Computation time: 3.839 seconds.
   user  system elapsed 
   2.89    0.37    7.89 

 

T-SQL query without any specifics

To have a baseline, let’s run the following query:

SET STATISTICS TIME ON;
SELECT 
[DofWeek]
,AVG(ArrDelay) AS [means]
FROM
    AirFlights_small
GROUP BY 
    [DofWeek]
SET STATISTICS TIME OFF;

And check these time statistics

 SQL Server Execution Times:
CPU time = 6124 ms,  elapsed time = 2019 ms.
Warning: Null value is eliminated by an aggregate or other SET operation.

Obiously the CPU / computation time is higher, although the elapsed time is faster.

ColumnStore Table

Let’s create a nonclustered column store index.

CREATE TABLE AirFlights_CS
(id INT IDENTITY(1,1)
,ArrDelay INT
,CRSDepTime DECIMAL(6,4)
,DofWeek NVARCHAR(20) 
);
GO
INSERT INTO AirFlights_CS(ArrDelay, CRSDepTime, DofWeek)
SELECT ArrDelay, CRSDepTime, DofWeek FROM AirFlights_small 

CREATE NONCLUSTERED COLUMNSTORE INDEX NCCI_AirFlight
ON AirFlights_CS
(id, ArrDelay, CRSDepTime, DofWeek);
GO

With the execution of the same query

SET STATISTICS TIME ON;
SELECT 
[DofWeek]
,AVG(ArrDelay) AS [means]
FROM
  AirFlights_CS
GROUP BY     [DofWeek] SET STATISTICS TIME OFF;

The following time statistics are in

 SQL Server Execution Times:
CPU time = 202 ms,  elapsed time = 109 ms.
Warning: Null value is eliminated by an aggregate or other SET operation.

 

In-Memory OLTP

To get Memory optimized table, we need to add a filegroup and create a table with memory optimized turned on:

CREATE TABLE dbo.AirFlight_M   
(  
  id INT NOT NULL PRIMARY KEY NONCLUSTERED
 ,ArrDelay INT
 ,CRSDepTime DECIMAL(6,4) 
 ,DofWeek NVARCHAR(20)
) WITH (MEMORY_OPTIMIZED=ON, DURABILITY = SCHEMA_AND_DATA);
GO

And insert the data

INSERT INTO AirFlight_M
SELECT * FROM AirFlights_small

Running the simple query

SET STATISTICS TIME ON;
SELECT 
[DofWeek]
,AVG(ArrDelay) AS [means]
FROM
    AirFlight_M
GROUP BY 
    [DofWeek]
SET STATISTICS TIME OFF;

results are:

 SQL Server Execution Times:
CPU time = 6186 ms,  elapsed time = 1627 ms.
Warning: Null value is eliminated by an aggregate or other SET operation.

These results were somehow expected, mostly because the ColumnStore table is the only one having index and reading (also by looking in execution plans) optimized with comparison to others. Also degree of parallelism, clustered and non-clustered index can  be pushed, but the idea was to have tests similar to the one in RevoScaleR and R environemnt. With R, we can not push any index on the XDF file.

In R we run:

system.time({ 
LMResults <- rxLinMod(ArrDelay ~ DayOfWeek, data = outFile2, transforms = transforms)
LMResults$coefficients
})

And in SSMS we run:

SET STATISTICS TIME ON;
-- 1. T-SQL
DECLARE @RScript nvarchar(max)
SET @RScript = N'library(RevoScaleR)
                LMResults <- rxLinMod(ArrDelay ~ DofWeek, data = InputDataSet)
                OutputDataSet <- data.frame(LMResults$coefficients)'
DECLARE @SQLScript nvarchar(max)
SET @SQLScript = N'SELECT ArrDelay, DofWeek FROM [dbo].[AirFlights_small]'
EXECUTE sp_execute_external_script
     @language = N'R'
    ,@script = @RScript
    ,@input_data_1 = @SQLScript
WITH RESULT SETS ((
            --DofWeek NVARCHAR(20)
        --    ,
            Coefficient DECIMAL(10,5)
            ));
GO
SET STATISTICS TIME OFF;


SET STATISTICS TIME ON;
-- 2. ColumnStore
DECLARE @RScript nvarchar(max)
SET @RScript = N'library(RevoScaleR)
                LMResults <- rxLinMod(ArrDelay ~ DofWeek, data = InputDataSet)
                OutputDataSet <- data.frame(LMResults$coefficients)'
DECLARE @SQLScript nvarchar(max)
SET @SQLScript = N'SELECT ArrDelay, DofWeek FROM [dbo].[AirFlights_CS]'
EXECUTE sp_execute_external_script
     @language = N'R'
    ,@script = @RScript
    ,@input_data_1 = @SQLScript
WITH RESULT SETS ((
            --DofWeek NVARCHAR(20)
        --    ,
            Coefficient DECIMAL(10,5)
            ));
GO
SET STATISTICS TIME OFF;


SET STATISTICS TIME ON;
-- 3. Memory optimized
DECLARE @RScript nvarchar(max)
SET @RScript = N'library(RevoScaleR)
                LMResults <- rxLinMod(ArrDelay ~ DofWeek, data = InputDataSet)
                OutputDataSet <- data.frame(LMResults$coefficients)'
DECLARE @SQLScript nvarchar(max)
SET @SQLScript = N'SELECT ArrDelay, DofWeek FROM [dbo].[AirFlight_M]'
EXECUTE sp_execute_external_script
     @language = N'R'
    ,@script = @RScript
    ,@input_data_1 = @SQLScript
WITH RESULT SETS ((
            --DofWeek NVARCHAR(20)
        --    ,
            Coefficient DECIMAL(10,5)
            ));
GO
SET STATISTICS TIME OFF;

 

Conclusion

Gathering statistics on CPU time and elapsed time when running simple Linear regression, this is comparison:

df_LR_comparison <- data.frame (
  method = c("T-SQL", "ColumnStore", "Memory Optimized", "RevoScaleR")
  ,CPUtime = c(3000,1625,2156,7689)
  ,ElapsedTime = c(14323,10851,10600,7760)
  )
library(ggplot2)

ggplot(df_LR_comparison, aes(method, fill=method)) + 
  geom_bar(aes(y=ElapsedTime), stat="identity") +
  geom_line(aes(y=CPUtime, group=1), color="white", size=3) +
  scale_colour_manual(" ", values=c("d1" = "blue", "d2" = "red"))+
  #scale_fill_manual("",values="red")+
  theme(legend.position="none")

Showing that elapsed time for R environment with RevoScaleR is fastest (and getting data from XDF), where as simple T-SQL run with sp_execute_external_script and using RevoScaleR gives the slowest response.

In terms of CPU time (white line), Columnstore with RevoScaleR call through external procedure outperforms all others.

Final conclusion: When running statistical analysis (using RevoScaleR or any other R library), use columnstore and index optimized tables/views to receive best CPU and elapsed times.  Important to remember is also the fact, that any aggregations and calculations that can be done within SQL Server, are better to be perfomered there.

 

As always, code is available at GitHub.

 

Happy coding! 🙂

Saving input and output with sp_execute_external_script using temporal table and file table (part #2)

In my previous blog post, Saving input and output with sp_execute_external_script, I was exploring the possibilities how to capture the R code that external procedure sends it internally to the Launchpad.exe program.  And a blog comment by reader Bob gave me additional push to write the second part to same topic. Thank you Bob for sharing this with us.

Bob was explained that how they are doing, and you can read all about it here. To recap, R code (as well as any additional packages) is stored on local file system, it gets copied, zipped, transferred and inserted to SQL table. Calling sp_execute_external_script everything gets unzipped and executed.

I will not comment on the solution Bob provided, since I don’t know how their infrastructure, roles, security is set up. At this point, I am grateful for his comment. But what I will comment, is that there is no straightforward way or any out-of-the-box solution. Furthermore, if your R code requires any additional packages, storing the packages with your R code is not that bad idea, regardless of traffic or disk overhead. And versioning the R code is something that is for sure needed.

To continue from previous post, getting or capturing R code, once it gets to Launchpad, is tricky. So storing R code it in a database table or on file system seems a better idea.

Starting with original sample R code:

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = N'
       d <- InputDataSet 
       c <- data.frame(Num_V1 = c(1,2,3))
       c
       OutputDataSet <- c'
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'

WITH RESULT SETS ((Numbers_From_R INT));

We can create SQL table for R code to be persistent and always available.

CREATE TABLE R_code 
(id INT
,R NVARCHAR(MAX))
INSERT INTO R_code
SELECT 1, '
       d <- InputDataSet 
       c <- data.frame(Num_V1 = c(1,2,3))
       c
       OutputDataSet <- c'
-- (1 row(s) affected)

DECLARE @r_code NVARCHAR(MAX)
SELECT @r_code = R FROM R_code WHERE id = 1

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = @r_code
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'
WITH RESULT SETS ((Numbers_From_R INT));

This will yield same results. To have R code and T-SQL code side by side, I would suggest to store T-SQL in table as well.

DROP TABLE IF EXISTS R_code
CREATE TABLE R_code 
(id INT
,R NVARCHAR(MAX)
,SQLC NVARCHAR(MAX))

INSERT INTO R_code
SELECT 1, '
       d <- InputDataSet 
       c <- data.frame(Num_V1 = c(1,2,3))
       c
       OutputDataSet <- c','SELECT 1 AS Nmbrs_From_R'
-- (1 row(s) affected)

DECLARE @r_code NVARCHAR(MAX)
DECLARE @sql_code NVARCHAR(MAX)
SELECT @r_code = R FROM R_code WHERE id = 1
SELECT @sql_code = SQLC FROM R_code WHERE id = 1

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = @r_code
    ,@input_data_1 = @sql_code
WITH RESULT SETS ((Numbers_From_R INT));

Now, the best thing to do, is to add some logging to the R_code table and some versioning. Easiest way to achieve this by using Temporal Table.

TEMPORAL TABLE

Rewrite the original table:

CREATE TABLE R_code 
(
 id INT IDENTITY(1,1)
,CombinationID INT NOT NULL CONSTRAINT PK_ComboID PRIMARY KEY
,R NVARCHAR(MAX)
,SQLC NVARCHAR(MAX)
,Valid_From DATETIME2 GENERATED ALWAYS AS ROW START NOT NULL
,Valid_To DATETIME2 GENERATED ALWAYS AS ROW END NOT NULL
,PERIOD FOR SYSTEM_TIME (Valid_From, Valid_To)
)
WITH (SYSTEM_VERSIONING = ON);

Please note, that table will be represented slightly differently (see the clock in the icon).

Besides actual table (or Temporal Table), system automatically creates history table where all the changes are being kept.

Once this is done, I can store T-SQL and R-code.

INSERT INTO R_code (CombinationID, R, SQLC)
SELECT 1,'
       d <- InputDataSet 
       c <- data.frame(Num_V1 = c(1,2,3))
       c
       OutputDataSet <- c','SELECT 1 AS Nmbrs_From_R'
-- (1 row(s) affected)

Now I can run the query same way as before:

DECLARE @r_code NVARCHAR(MAX)
DECLARE @sql_code NVARCHAR(MAX)
SELECT @r_code = R FROM R_code WHERE CombinationID = 1
SELECT @sql_code = SQLC FROM R_code WHERE CombinationID = 1

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = @r_code
    ,@input_data_1 = @sql_code
WITH RESULT SETS ((Numbers_From_R INT));

Suppose that there are changes either to R code or SQL Code committed, and I will simulate this change with an UPDATE query:

-- INSERT ANOTHER CombinationID = 1, with changed R Code
UPDATE R_code
SET R = '
       d <- InputDataSet 
       c <- data.frame(Num_V1 = c(1,2,3,4))
       c
       d
       OutputDataSet <- c'
,SQLC = 'SELECT 1 AS Nmbrs_From_R'
WHERE
    CombinationID = 1
-- (1 row(s) affected)

The best part is that system maintains the versioning automatically and I – as an end user – don’t need to worry about changing the original T-SQL code that executes R Script. So once again I can execute the same query:

DECLARE @r_code NVARCHAR(MAX)
DECLARE @sql_code NVARCHAR(MAX)
SELECT @r_code = R FROM R_code WHERE CombinationID = 1
SELECT @sql_code = SQLC FROM R_code WHERE CombinationID = 1

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = @r_code
    ,@input_data_1 = @sql_code
WITH RESULT SETS ((Numbers_From_R INT));

And now, I will get the results from updated R-script or T-SQL code.

FILE TABLE

Another way (among many) is to use File Table. In one of my previous blog posts, I covered how to create and configure File Table. In this scenario, we will consider following. Create a file, that will contain R code and store it with *.R extension. And upload it to the directory, where File Table is considering this file-stream data to be landed or stored.

Quickly check the configuration:

--- Check configurations
SELECT 
  DB_NAME(database_id) AS DbName
 ,non_transacted_access
 ,non_transacted_access_desc
 ,directory_name  
 ,*
FROM  sys.database_filestream_options
WHERE 
    DB_NAME(database_id) = db_name() --'FileTableRChart'

By checking where my R file is residing:

SELECT 
     FT.Name AS [File Name]
    ,IIF(FT.is_directory=1,'Directory','Files') AS [File Category]
    ,FT.file_type AS [File Type]
    ,(FT.cached_file_size)/1024.0 AS [File Size (KB)]
    ,FT.creation_time AS [File Created Time]
    ,FT.file_stream.GetFileNamespacePath(1,0) AS [File Path]
    ,ISNULL(PT.file_stream.GetFileNamespacePath(1,0),'Root Directory') AS [Parent Path]
FROM 
    [dbo].[ChartsR] AS FT
LEFT JOIN [dbo].[ChartsR] AS PT
ON FT.path_locator.GetAncestor(1) = PT.path_locator
WHERE
    FT.File_type = 'R'

And you can see that I have created R file with the name R_combination1.R

So we can access this using OPENROWSET.

SELECT * FROM 
OPENROWSET(BULK N'\\****\RCharts\DocumentTable\R_Combination1.R',
   SINGLE_CLOB) AS R_Code

You can also do this using the master.dbo.xp_cmdshell, whatever suits you better.

Once you have code read from R file, you can simply continue to execute external procedure.

-- Physical Location of FileTable
DECLARE @r_code NVARCHAR(MAX)
SELECT @r_code = BulkColumn FROM 
OPENROWSET(BULK N'C:\DataTK\00\R_Combination1.R', SINGLE_CLOB) AS R_Code

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = @r_code
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'
WITH RESULT SETS ((Numbers_From_R INT));

And result is the same. In this case you need to consider the usage of BCP, BulkCopy or XP_CMDSHELL. Again, based on your eco-system, what suits you best.

With file table versioning can be a bit of work-around. In addition, you should implement some naming convention to store files correctly or add additional logic to storing and keeping the changes.

So in my case, I am keeping versioning on FileName level, which can be done using T-SQL or renaming the file on file system and later creating hierarchies with files.

 

As always, Code is available at GitHub.

 

Happy coding!

Saving input and output with sp_execute_external_script

Again I was at the point, where I needed to store and save to external file all the R code that was executed through sp_execute_external_script.

Soon, you will find out several interesting things. To show the example, I will start with following example:

USE [WideWorldImporters];
GO

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = N'
       d <- InputDataSet 
       c <- data.frame(Num_V1 = c(1,2,3))
       c
       OutputDataSet <- c'
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'

WITH RESULT SETS ((Numbers_From_R INT));

The result is a column called “Numbers” with three rows, represented from the data frame. This is very easy and straight-forward.

DMV

By using dynamic management view sys.dm_exec_query_stats as following:

SELECT
     QM_ST.[TEXT] AS [Query]
    ,DM_QS.last_execution_time
    ,DM_QS.query_hash
    ,DM_QS.query_plan_hash
 FROM 
    sys.dm_exec_query_stats AS DM_QS
    CROSS APPLY sys.dm_exec_sql_text(DM_QS.sql_handle) AS QM_ST
ORDER BY 
    DM_QS.last_execution_time DESC

Surprisingly I get only the following query returned:

sp_execute_external_script: SELECT 1 AS Nmbrs_From_R

which is far what was executed in the first place!

EXECUTION PLANS

When using sys.dm_exec_query_plan dynamic management view to generate executed query plan, I get similar result with no R code and little sign of SQL query that was introduced to sp_execute_external_query procedure.

Relative the same results emerges when showing actual execution plan in SSMS. Only XML-UDX is showed.

So far, very slim possibility to get some extra and additional information from query statistics DMV or execution plan.

SQL SERVER PROFILER

So opening SQL Profiler and running the example sp_execute_external_script code, I was finally able to see the actual R code within profiler:

Upon closer look, we can see that profiler wraps execution of external procedure with following command SET STATISTICS XML ON/OFF. So we can store the results from profiler into a table or trace file and later filter out the R-code!

QUERY STORE

Query store is very very useful and new feature with flagship MSSQL2016. Storing the queries and execution times is therefore needed in order to do later performance analysis. So in this phase, let’s just see, if we can store external procedure code in query store.

With execution of R external procedure, I execute following query to check the Query Store (QS):

SELECT 
  QSQT.query_text_id
 ,QSQT.query_sql_text
 ,QSP.plan_id
FROM 
    sys.query_store_plan AS QSP
    JOIN sys.query_store_query AS QSQ  
    ON QSP.query_id = QSQ.query_id  
    JOIN sys.query_store_query_text AS QSQT  
    ON QSQ.query_text_id = QSQT.query_text_id

And the results are – in a way – not surprising at all, since many of query store statistics base on DMV. So result for my external procedure is again, very little informative in order to extract R code:

Something, we have seen already couple of times. And no sign of execution of R Script. In fact, looking from this, it is hard even to tell, this was passed to RLaunchpad.exe external program.

SINK

Sink is a R function to store the output of the executed R code into external file. With execution of any of the two T-SQL code, I will never be able to either get the results nor the R code itself.

In case of results:

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = N'
        sink("C:\\DataTK\\logRSQLsession3.txt")
        d <- InputDataSet
        c <- data.frame(Num_V1 = c(1,2,3))
        c
        sink()
        OutputDataSet <- c'
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'
WITH RESULT SETS ((Numbers_From_R INT));

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = N'
        c <- data.frame(Num_V1 = c(1,2,3))
        c
        sink("C:\\DataTK\\logRSQLsession3.txt")'
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'
WITH RESULT SETS NONE;

In both cases the file is created, but it is just that. Empty file. No content whatsoever.

LOAD

Load will store intermediate results into file for later analysis or for semi aggreagated data, used for further calculations. So, I have tested it as following:

EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = N'
        c <- data.frame(Num_V1 = c(1,2,3))
        c
        save(c, file="C:\\DataTK\\logRSQLsession3.rda")
        #load(file="C:\\DataTK\\logRSQLsession3.rda")'
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'
WITH RESULT SETS NONE;


-- LOAD RESULTS
EXEC sys.sp_execute_external_script
     @language = N'R'
    ,@script = N'
        load(file="C:\\DataTK\\logRSQLsession3.rda")
        OutputDataSet <- c'
    ,@input_data_1 = N'SELECT 1 AS Nmbrs_From_R'
WITH RESULT SETS ((Num_V1 INT));

 

EXTENSIBILITY LOG

Extensibility Log will store information about the session but it will not store the R or R environment information or data, just session information and data. Navigate to:

C:\Program Files\Microsoft SQL Server\MSSQL13.MSSQLSERVER\MSSQL\LOG\ExtensibilityLog

to check the content and to see, if there is anything useful for your needs.

Conclusion

We are very limited in terms of exporting executed R code, results or Logs. Same applies for importing any additional code. We have seen that import, source are not working, whereas Load for loading *.rda files is working. At least something 🙂 There should be more ways to get into the, especially with Rterm or Vanilla R, but the idea was to have everything run comfortably from the SSMS environment.

As you can see, there is little possibilities to store R code separately or store execution R logs in external files. But I presume, I haven’t exhausted all the possibilities, so there should be still some ways to try and do this.

As always, the code is available at Github.

Happy Rrrrr!

Linear regression in “The Man who counted”

Recently, I got a book by Brasilian writer  Júlio César de Mello e Souza (published under pen name Malba Tahan), titled The Man who counted. Book is a collection of mathematical stories very similar to Scheherazada’s 1001 Nights, where mathematical story-telling is the center of book.

                                              

In story 5“In so many words”, Malba describes a simple algebraic problem of proportion between price for lodging offered and price of jewel sold.

This man,” old Salim said pointing to the jeweler “came from Syria to sell 
precious stones in Baghdad. He promised he would pay 20 dinars for his 
lodgings if he sold all of his jewels for 100 dinars and 35 dinars if he 
sold them for 200. After several days of wandering about, he ended up selling 
all of them for 140 dinars. How much does he owe me according to our agreement?”

Both, jeweler and lodge owner,  calculate the result each using percent proportion problem, both ending with wrong results, that was each to favor each:

1. Jeweler:

200 : 35 :: 140 : x

x = (35 x 140) / 200 = 24.5

2. Lodge Owner:

100 : 20 :: 140 : x

x = (20 x 140) / 100 = 28

 

With two different results, both would end up in argument, so the third needed to be calculated:

Sale Price        Price of Lodgings
200               35
-100             -20
-----            -------
100               15

 

So the difference between both calculations forms a proportion to calculate the new case, when Sale price for jewel is 140, the price of lodging would be 26.

100: 15:: 40: x

x = (15 x 40) / 100 = 6

 

Mathematically speaking, problem is very interesting to be solved also using Linear Regression, since the two pair of points [200, 35] and [100, 20] form a linear prediction function and we would need to predict what would be the price of lodging, when sale price for jewel is 140.

diamond <- c(100, 200)
sleep   <- c(20, 35)

# regression
sleep_model <- lm(sleep ~ diamond)

plot(x=diamond, y=sleep)
abline(lm(sleep ~ diamond))

Now, we can call this a prediction, what actually Beremiz does by heart.

predict_data <- data.frame(diamond=140)
fit <- predict(sleep_model, predict_data, interval = "predict")

#new value for diamond=140
fit[1]

Result is 26, which is strictly algebraic and R-prediction speaking correct result.

In this case, linear regression does same as proportion calculation, but what strikes me is which calculation – not mathematically speaking – does make more sense? 26 or 24,5 or 28 ? And which method for calculating next price lodge should satisfy both jeweler and lodge owner.

Happy reading!

 

 

Is it possible to use RevoScaleR package in Power BI?

I was invited to deliver a session for Belgium User Group on SQL Server and R integration. After the session – which we did online using web based Citrix  – I got an interesting question: “Is it possible to use RevoScaleR performance computational functions within Power BI?“. My first answer was,  a sceptical yes. But I said, that I haven’t used it in this manner yet and that there might be some limitations.

The idea of having the scalable environment and the parallel computational package with all the predictive analytical functions in Power BI is absolutely great. But something tells me, that it will not be that straight forward.

So let’s start by taking a large (500 MB) txt file and create XDF file:

library(RevoScaleR)
file.name <- "YearPredictionMSD.txt"
rxOptions(sampleDataDir = "C:\\Files")
sampleDataDir

File is available on-line at this address with the zip file.

Getting data with R script

Open Power BI and choose Get Data -> R Script -> and copy/Paste the following slightly changed code:

library(RevoScaleR)
file.name <- "YearPredictionMSD.txt";
rxOptions(sampleDataDir = "C:\\Files");
sampleDataDir

With copy pasting and clicking OK,

You will have to wait for the data to be read into the memory, the data models to be created and after monitoring the memory consumption and patiently waiting, you will notice, that this particular dataset (500 MB or 160 MB XDF), that minimum 3 GB of RAM will be consumed and you will end up with preview:

By now, you will also notice that after saving this Power BI document, it will take somewhere up to 700 MB of your disk space and all the data visualization will consume additional RAM and time. After you will close the Power BI document, you will notice a lot of RAM being released.

Using R Script in the visuals

When you create a new Power BI document, I will create new dataset by Entering data. I will create three “dummy” variables.

With these three variables I will try to inject the data returned from XDF data format and have data represented in Power BI.

After selecting the new visual and choosing R visual, I inserted following code:

library(RevoScaleR)
file.name <- "YearPredictionMSD.txt";
rxOptions(sampleDataDir = "C:\\Files");
sampleDataDir

And this time, the result is fascinating. R is plotting histogram in a split of a second, simply meaning it takes advantage of XDF file and inject it to Power BI.

This is still – an outer file or dataset -, that Power BI does not have a clue about. Meaning, no slicers are available for dynamic change of the user selection.

Let’s try to insert the data into those three dummy variables, where the third one will be a factor that I have to pre-prepare. Since in this case factor is Year, it is relatively easy to do:

library(RevoScaleR)
library(gridExtra)
library(dplyr)
Year % filter(year == c("2000","2001","2002")))
grid.table(df_f %>% filter(year == Year))

Once I have this inserted in new R visualize, I just need to add a dummy slicer.

Now, I can easily change the years for my cross-tabulation (using rxCrosstab function). Since calculation is comprehended in the back on the whole dataset and using dplyr package just to omit or filter the results, it is also possible to use rxDatastep:

rxDataStep(inData=outputFile, outFile="C:\\Files\\YearPredictMSD_Year.xdf", 
             overwrite=TRUE, transforms=list(LateYears = V1 > 1999))
rxCrossTabs(V2~F(LateYears), data = "C:\\Files\\YearPredictMSD_Year.xdf")

In this way, you will be creating new XDF file through PowerBI with the transformation. Bear in mind, that this step might take some extra seconds to create new variable or to make a subset, if you would need. Again, this is up to  you to decide, based on the file size.

Using SQL Server procedure with R Script

This approach is not that uncommon, because it has been proven that using Stored Procedures with T-SQL and R code is useful and powerful way to use SQL Server and R integration within SSRS.  Changing the computational context is sure another way to make a work around.

Creating Stored procedure:

CREATE PROCEDURE [dbo].[SP_YearMSD_CrossTab]
AS
BEGIN
    DECLARE @RScript nvarchar(max)
        SET @RScript = N'
                library(RevoScaleR)
                sampleDataDir

Or by copying the T-SQL Code into the SQL Server Data Source, the result is the same.

In both cases, you should have a cross-tabulational  representation of XDF dataset within Power BI. And now you can really use all the advantages of Power BI visuals, Slicers and as well any additional R predictions.

There is a slight minus to this (if not all) approaches like this. You need to have many stored procedures or queries having generated like this. Also rxCube will help you to some extent, but repetitive work will not be avoided.

Using HDInsight or Hadoop?

Using XDF data files stored in HD-Insight or in Hadoop would generaly mean using same dataset and step as for SQL Server procedure. Just that you would need to – prior to executing T-SQL script, also change comptutational context:

# HD Insight - Spark - Azure
HDInsight mySshUsername = USNM,mySshHostname = HSTNM,
mySshSwitches= SWTCH) 
rxSetComputeContext("HDInsight")
## Hadoop
Hadoop mySshUsername = USNM,mySshHostname = HSTNM,
mySshSwitches= SWTCH)
rxSetComputeContext("Hadoop")

Verdict

I have explored couple of ways how to use the Power BI visuals and environment with RevoScaleR XDF (eXternal Data Frame) datafiles. I have to admit, I was surprised that there will be a way to do it in a relatively easy way, but from data scientist perspective, it is still some additional load and work before you can start with actual data analysis. Last two approaches (R script in Visuals and SQL Server Procedures) are by far the fastest and also take the advantage of using parallel and distributed computations that RevoScaleR package brings.

I would very strongly advise Microsoft and Power BI development team to add XDF plug-in to Power BI. Plug-in would work with metadata presentation of the data each time the computations should be used, the metadata would push the code against R Server to have results returned. This would, for sure be a great way to bring Big Data concept to Power BI Desktop.

As always, code and samples are available at GitHub.

Happy coding!