data management using r – interfacing with the structured query language

24
© 2003 By Default! A Free sample background from www.powerpointbackgrounds.com Slide 1 Data Management Using R – Data Management Using R – Interfacing with the Interfacing with the Structured Query Language Structured Query Language STAT 7550 – Statistical Computing Utah State University November 21, 2008 Bill Welbourn

Upload: kim-johns

Post on 02-Jan-2016

21 views

Category:

Documents


1 download

DESCRIPTION

Data Management Using R – Interfacing with the Structured Query Language. STAT 7550 – Statistical Computing Utah State University November 21, 2008 Bill Welbourn. Objectives of the Project. Introduce the notion of the database. Database applications. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 1

Data Management Using R – Data Management Using R – Interfacing with the Structured Interfacing with the Structured

Query LanguageQuery Language

STAT 7550 – Statistical ComputingUtah State University

November 21, 2008Bill Welbourn

Page 2: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 2

Objectives of the Project Objectives of the Project • Introduce the notion of the database.

• Database applications.

• General overview of the SQLite Relational Database Management System (RDBMS).

• Explain how R1 (and other programming languages) interfaces with the SQLite RDBMS. Highlights of the R commands for the interface to the SQLite RDBMS, the RSQLite library.

• A working example, demonstrating the procedure for storing and retrieving an R dataframe within a SQLite database.

• Further motivation for the use of the R-SQLite interface, working with “massive” databases.

1R Development Core Team (2008), Version 2.8.0.

Page 3: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 3

What is a Database?What is a Database?• Essentially a series of structured files on a

computer that are organized in a highly efficient manner.

• The organization is comprised in a hierarchical manner, from the “top, down,” as shown in Figure 1 below.Figure 1: The anatomy of a database

Database

TableColumn Column

Row Field FieldRow Field Field

TableColumn Column

Row Field FieldRow Field Field

TableColumn Column

Row Field FieldRow Field Field

Page 4: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 4

Components of a TableComponents of a Table• As Figure 1 suggests, at the highest level, a database is

comprised of a series of tables.

• Each table is made up of a series of columns. Think of the columns as characteristics (variables) collected for a study.

• Data is stored in rows of the table, where each row of the table is called a record. Records of a table are essentially synonymous with observations for a study.

• The location where each row intersects a column is known as a field.

• Each table contains specific, common data. A table of a database is analogous to a worksheet within an Excel workbook.

Page 5: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 5

What is a Relational What is a Relational Database?Database?• It is a database comprised of tables which relate to one

another. The table relationships are based on “Key fields.”

• To illustrate, consider two relational tables within a database. A column of each table is affixed with the same naming convention, say “ID.” Each field within these columns is designated a unique (key) label, so that a one-to-one (relationship) mapping between the tables is obtained.

Table i

ID Column 2 … Column k

Row 1 Key 1 Fieldi (1,2) … Fieldi (1,k)

Row 2 Key 2 Fieldi (2,2) … Fieldi (2,k)

… … … … …

Row n Key n Fieldi (n,2) … Fieldi (n,k)

Table j

Column 1 ID … Column m

Row 1 Fieldj (1,1) Key 1 … Fieldj (1,m)

Row 2 Fieldj (2,1) Key 2 … Fieldj (2,m)

… … … … …

Row n Fieldj (n,1) Key n … Fieldj (n,m)

Figure 2: Example of two relational tables

Page 6: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 6

Three Types of RelationshipsThree Types of Relationships• One-to-One

• A record in table one must have a record in table two, and vice-versa. Example: More variables (columns) were collected in a study than allowed to be stored in a single database table. Study participants (observations/records) are labeled with unique ID’s which allow for the table-to-table relationship to be established.

• One-to-Many• A record in table one has many corresponding records in table two,

while table two has many records which correspond to a single record in table one. Example: Study participant identifiers (unique ID’s) are stored in table one, while repeated measurements are stored in table two.

• Many-to-Many• Like the one-to-many relationship, table one has many

corresponding records in table two. However, unlike the one-to-many relationship, table two has many corresponding records in table one. Example: Customer product orders. Each order can contain multiple products, and one product can be in many orders.

Page 7: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 7

Database ApplicationsDatabase Applications• World Wide Web.

• Medical Data.

• Data analysis situations which warrant the consideration in utilizing a database:

• You possess a flat (text) file(s) with an inordinate number of observations.

• You have collected an insurmountable quantity of characteristics for your observations (e.g., genetic data).

• You are ready to execute a large simulation analysis.

• You need to prepare a portable file, so that another statistician has easy access to your data.

• Anytime there is data in your possession. It is fairly straightforward to maintain a SQL database in R.

Page 8: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 8

The SQLite RDBMSThe SQLite RDBMS• Created by D. Richard Hipp. Version 1.0 released

August 17, 2000. Most recent version, 3.6.4, releasedOctober 15, 2008.

• An ACID (Atomicity, Consistency, Isolation, Durability) compliant RDBMS. In computer science, ACID is a set of properties which guarantee that database transactions (logical operations) are processed reliably.

• Contained in a relatively small (~500kB) C programming library.

• It is not a database, rather a system which manages databases. Microsoft Access, in contrast, is simply a program used to create a database.

Page 9: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 9

The SQLite RDBMS (cont.)The SQLite RDBMS (cont.)• The columns of a table for a SQL database,

typically are assigned a “type” (e.g., string, integer, float, double). This is analogous to defining variables in the C programming language. However, SQLite (automatically) assigns types to individual values.

• Allows for multithread reading of a database. The writing of a database can only occur if no other access to the database is present.

• Interfacing with programming languages (e.g., BASIC, C, C++, Perl, Ruby, and R).

• Most widely deployed SQL RDBMS.

Page 10: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 10

Interfacing the SQLite RDBMSInterfacing the SQLite RDBMS• Consists of three components: The application

(such as R) which requires access to the database; an interface; and the RDBMS.

• An interface acts as an interpreter, translating commands from the application, so that the database is accessible to the user. In R, the interface lies within the DBI library.

• The interface communicates with the database via the applicable database driver. The database driver knows how to “talk” to the database. In R, the SQLite database driver and the source (C library) for the SQLite engine are included within the RSQLite library.

Page 11: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 11

The R-SQLite InterfaceThe R-SQLite InterfaceFigure 3: The process flow between the application and

the database

Application (e.g., R)

Interface (e.g.,DBI library in R)

Database

Driver (e.g., SQLite in R)

Page 12: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 12

Accessing a SQLite DBAccessing a SQLite DB• A five step cycle:

1) Connect to the database. In R, to establish the connection to a database, issue the commands: dbDriver(); and dbConnect().

2) Issue a query or command to the database. To issue a query in R, the command, dbSendQuery(), is (typically) used. Queries consist of SQL commands.

setwd("c:/SQL"); library(DBI); library(RSQLite)

dbfile<-"DATA.dbsql"; drv<-dbDriver("SQLite")con<-dbConnect(drv,dbname = dbfile)

rs <- dbSendQuery(con, "select v1,v2,v3 from Table1 where v1==1")rs<-dbSendQuery(con,"select * from Table1")

Brief Summary of SQL Commands

SQL Command Tabular Parameters Required of SQL Command

Select Column Label(s)

From Table Name(s)

Where Specific Values for Column Label(s)

Order by Column Label(s)

Page 13: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 13

Accessing a SQLite DB (cont.)Accessing a SQLite DB (cont.)3) If a query was issued, we need to retrieve the applicable

recordset. To do this in R, we use the command, fetch(). The recordset will exist as a dataframe in R.

4) Clear the query result, manipulate the recordset, and update the database. To clear a query in R, use the command, dbClearResult(). To update a database, issue the R command, dbWriteTable().

5) Close the connection to the database. To do this in R, use the command, dbDisconnect().

d1<-fetch(rs, n = -1)

dbDisconnect(con)

dbClearResult(rs)dbWriteTable(con, “Table”, data frame, append, row.names, overwrite)

Page 14: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 14

Example 1Example 1• You have recorded n (unique) observations (records)

for a study, and collected m-1 characteristics (excluding the unique observation identifier) for each observation. Having the option to store your data as a flat file or as a (single table) SQL database, which should you choose?

• To address this issue, you decide to conduct a (small) simulation analysis, investigating data retrieval times for the two types of data repositories. Figure 4, shown on the subsequent slide, displays the results from a simulation, where m=50 and each (of the 49) characteristic is of type “double.” A total of 100 distinct values of n, n1,…,n100, were chosen, in accordance to the rule

Page 15: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 15

Example 1 (cont.)Example 1 (cont.)Figure 4: Flat file – DB comparison

Page 16: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 16

Example 2Example 2• You have recorded data for n (unique) participants of a

study, and collected m-1 characteristics (excluding the unique observation identifier) for each participant. Further, for the ith participant, you have recorded a total of i record(s). Having the option to store your data as a flat file or as a (single table) SQL database, which should you choose? Given the unique identifier for a participant, suppose it is desirable to have quick access to the records for each participant.

• Figures 5 and 6, shown on the subsequent slides, display the results of data retrieval times, where n=200 and n=500, respectively, m=50, where each variable type is “double”. The displayed value for the vertical axis, is the required time to read in the i records for the ith participant.

Page 17: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 17

Example 2 (cont.)Example 2 (cont.)Figure 5: Flat file – DB comparison, marginal

read I

Page 18: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 18

Example 2 (cont.)Example 2 (cont.)Figure 6: Flat file – DB comparison, marginal

read II

Page 19: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 19

Example 3Example 3• You have recorded data for n (unique) participants of a

study, and collected allele types at more than two million (2x106) single nucleotide polymophism (SNP) sites in the human nuclear genome. How could we utilize a relational database to represent the repository for these data?

• It turns out that a SQLite database table is limited to 999 columns. So, we simply create a sufficient number of tables (each with say m columns), and populate the tabular columns with the SNP data, making sure to create a “Key” column for each table.

• Figures 7 and 8, display the required time (by database table) to retrieve the first two columns – the “Key” along with a column of data – for SQL databases of size n=2,500 and n=25,000, respectively. For each database, a total of 2,106 tables were created, where m=951 columns for each database table. That is, these two databases, comprise slightly greater than five billion and 50 billion fields, respectively.

Page 20: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 20

Example 3 (cont.)Example 3 (cont.)Figure 7: Retrieval Time for a Massive SQL

Database I

Page 21: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 21

Example 3 (cont.)Example 3 (cont.)Figure 8: Retrieval Time for a Massive SQL

Database II

Page 22: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 22

Example 3 (cont.)Example 3 (cont.)Figure 9: Retrieval Time for a Massive SQL

Database III

Page 23: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 23

ConclusionConclusion• Database advantages

• Ability to store an extraordinary quantity of data. On a Windows NT platform, a SQL table can be as large as 2TB; on 64-bit operating systems, there is virtually no limit to the size of a SQL table.

• A single file could be utilized as a central data warehouse.

• Ability to create tabular relationships.

• Database indexing makes for very fast data retrieval.

• Portability.

• Interfacing with programming languages.

• Multithreading.

• Database disadvantages• Can be a bit of a challenge to recall, “What data, lives in which

table?”

• There is no “safety net” when it comes to overwriting data in a database.

• Essentially having to learn a new programming (querying) language.

• Database administration is industry requires continuous careful maintenance.

Page 24: Data Management Using R – Interfacing with the Structured Query Language

© 2003 By Default!

A Free sample background from www.powerpointbackgrounds.com

Slide 24

References Hogan, R (2002). A practical guide to database design. Prentice Hall, Englewood Cliffs, NJ.

• This is a good resource to obtain a working knowledge of what a database is all about. It covers the issues (e.g., who will use the database, and what should the database contain), say an employer, would consider prior to implementing a database in practice. It does not, however, discuss how to create and maintain the database, from a programming point of view (I.e., the SQL commands).

Maslakowski M, Butcher T (2000). SAMS teach yourself MySQL in 21 days. Macmillan USA, Indianapolis, IN

•Although the Windows version of R does not have the RMySQL binary package, this book is an excellent resource to “getting your feet wet” with the SQL programming language. The book lacks the theme of what the Hogan (2002) book comprises. Namely, the source does not talk about strategies in database design.

R Special Interest Group on Databases (R-SIG-DB). A common database interface (DBI). Software version 0.2-4 retrieved from http://cran.r-project.org/, September 27, 2008.

•The DBI package is necessary (but not sufficient) to interface any database with R (see slide 11). The DBI.pdf (available at the web link provided above) is a good document to review, prior to creating your first database in R. It provides an overview of the DBI package, and like most programming in R, to learn (effective) database management in R will require practice through working with data frames.