Special Interest Activity Special Interest Activity Trio – A System for Integrated Management of Data, Uncertainty, and Lineage ITK478 Yan Cui
Jan 14, 2016
Special Interest ActivitySpecial Interest ActivityTrio – A System for Integrated Management of Data, Uncertainty, and Lineage
ITK478Yan Cui
AgendaAgendaUnderstand Trio new database systemSystem requirements of installing Trio systemSource codes/packages for Trio systemProcedure of Trio system installationConfiguration of Trio SystemExperiment Trio DBMS Using TrioExplorer and
TrioPlus (demo)Trio API Integrated Into Other Python ScriptsTrio API and Translator (Python)Trio Query Language (TriQL) Structure (demo)Advantage and Disadvantage of using Trio
DBMSDocumenting Trio system report/bugs
Understand Trio new database Understand Trio new database systemsystemBasic concept of new Trio DBMS
◦ Trio is a new kind of database system (DBMS), which was developed by Stanford University Lab at Dec, 2006.
◦ An extended relational model called Uncertainty Lineage Database (ULDB) , and also supports Trio’s query language Called TriQL [1].
◦ Handles structured data, uncertainty of data, and data lineage together in a fully integrated manner.
Trio System architecture (components)Uncertainty-Lineage Database (ULDB)TriQL: The Trio Query Language
Trio System architectureTrio System architecture
Four primary components of Trio DBMS◦ command-line client (TrioPlus)◦ TrioExplorer◦ Trio API and translator (Python)◦ standard relational DBMS (PostgreSQL)
Uncertainty-Lineage of Database◦ Encoded data table◦ Lineage table◦ Trio metadata◦ Trio Stored Procedures
Uncertainty-Lineage Database Uncertainty-Lineage Database (ULDB)(ULDB)Alternatives‘?’ (Maybe) AnnotationsNumerical ConfidencesLineageSample: Drives (person, color, car) and Saw
(witness, color, car) uncertainty tables with/no with confidence.
Uncertainty-Lineage Database Uncertainty-Lineage Database (ULDB)(ULDB)Alternatives
◦ Definition: Alternatives are presenting uncertainty about the contents of a tuple [2]
◦ ‘||’ annotation◦ Drives(person, color, car) and Saw (witness, color,
car) uncertainty tables◦ ‘Select * from Drives”
Uncertainty-Lineage Database Uncertainty-Lineage Database (ULDB)(ULDB)‘?’ (Maybe) Annotations
◦ Definition: ‘?’ annotation present the existence of a tuple on the x-tuple, also called maybe x-tuple [2]
◦ Drives(person, color, car) and Saw (witness, color, car) uncertainty tables
◦ ‘select * from Drives’
Uncertainty-Lineage Database Uncertainty-Lineage Database (ULDB)(ULDB)Numerical Confidences
◦ Definition: Numerical confidence also was considered as probability [2].
◦ Drives(person, color, car) and Saw (witness, color, car) uncertainty tables with confidence
◦ ‘select * from Drives’
Uncertainty-Lineage Database Uncertainty-Lineage Database (ULDB)(ULDB)Lineage
◦ Definition: “recorded at the granularity of tuple alternatives: Lineage connects as x-tuple alternative to other x-tuple alternative.” in [2]
◦ Drives(person, color, car) and Saw (witness, color, car) uncertainty tables with confidence
◦ ‘select person from Drives’
TriQL: The Trio Query LanguageTriQL: The Trio Query Language
Two major parts of TriQL◦ built-in functions and predicates for querying
confidence values and lineage◦ regular SQL syntax
System requirements of System requirements of installationinstallationOperating systems - as Linux, Mac OS X, and
Win-32 (XP, Vista, and 32-bit Server)
PostgreSQL database (version 8.2.5, 8.1.10, 8.0.14, and 7.4.18) - Linux and Win32
Python API - windows, Linux/Unix, Mac OS X, OS/2, and Amiga
Source codes/packages for Trio Source codes/packages for Trio systemsystemListing source codes Python 2.4 can be downloaded from http://www.python.org/ . Easy_install can be downloaded from
http://peak.telecommunity.com/DevCenter/EasyInstall and the file called ez_setup.py.
Readline 1.7.win32 can be downloaded from http://www.python.org/ .
ctypes-1.0.2.win32-py2.4 can be download from http://www.python.org/ .
PostgreSQL 8.1 can be downloaded from http://www.postgresql.org/
Graphviz 2.14 is the only version compatible with Trio API. It is available in http://infolab.stanford.edu/trio/code/graphviz-2.14.1.exe .
PyGreSQL can be downloaded from http://www.pygresql.org/ . Pylons 0.9.5 can be downloaded from http://pylonshq.com/ . PLY 2.2 can be downloaded from http://www.dabeaz.com/ply/
. PyParsing can be downloaded from
http://pyparsing.wikispaces.com/ . PyDot can be downloaded from
http://code.google.com/p/pydot/downloads/list . Trio API 1.0 can be downloaded from
http://infolab.stanford.edu/~theobald/sources/TRIO.zip .
Procedure of Trio system Procedure of Trio system installationinstallationPython
◦ Python 2.4 windows version (python-2.4.4.msi).◦ Install Python in C:/Python directory◦ Set path=c:/Python24; in environment variables
Readline◦ Readline-1.7.win32-py2.4.exe and install into Python
directoryCtypes
◦ Ctypes-1.0.2.win32-py2.4.exe and install into Python directory
PostgreSQL◦ PostgreSQL 8.1 windows version (postgresql-8.1.msi )◦ Install PostgreSQL 8.1 and set path C:\Program Files\
PostgreSQL\8.1\bin; after completed the installation.Graphviz
◦ Graphviz 2.14 version, install to your workstation and set path C:\PROGRA~1\ATT\Graphviz\bin; in environment variables after completed the installation
Procedure of Trio system Procedure of Trio system installation (cont)installation (cont)Easy_install - Download ez_setup.py in C:/ directoryPyGreSQL – In command line, cd\ to c: directory, and
run python ez_setup.py PyGreSQL to install components.
Pylons – In command line, cd\ to c: directory, and run python ez_setup.py Pylons==0.9.5 to install Pylons. Set path c:\python24\Scripts in environment variables.
PLY – In command line, cd\ to c: directory, and run python ez_setup.py Ply==2.2 or easy_install Ply==2.2.
PyParsing – In command line, cd\ to c: directory, and run python ez_setup.py PyParsing.
PyDot – download the source from website. Access to folder in command line, and then install manually by running ‘python setup.py install’.
Trio API◦ Download source code in any directory◦ Copy Trio-1.0\spi\triospi_win32.dll to PostgreSQL’s lib
directory and renamed as triospi.dll
Procedure of Trio system Procedure of Trio system installation (cont)installation (cont)
PostgreSQL installation
Configuration of Trio SystemConfiguration of Trio System
Windows superuser authentication to access PostgreSQL
TrioExplorerTrioPlus
Configuration of Trio SystemConfiguration of Trio SystemWindows superuser authentication for
PostgreSQL◦ Open pgadmin III for PostgreSQL◦ Create new login role
Role name ‘myname’ (as same as windows login account)
Set password (Password can be any) and check all role privileges and click ok.
◦ Create new database Database name ‘myname’ (as same as username) Owner is ‘myname’ and click ok
◦ Initialize Trio schema information In Trio-1.0\setup, open setup.py with notepad to
comment out the last three codes and put the following. os.system("psql %s %s < setup.sql" % (pgdbname,
username))os.system("psql %s %s < setup_triospi.sql" %
(pgdbname, username)) os.system("psql %s %s < trio_get_conf.sql" %
(pgdbname, username)) Save the file, and at the command line, cd \Trio-1.0\
setup, and run ‘python setup.py myname myname’
Configuration of Trio System Configuration of Trio System (cont)(cont)pgadmin III for PostgreSQL 8.1
Configuration of Trio System Configuration of Trio System (cont)(cont)TrioExplorer
◦ Make sure PostgreSQL is working.◦ Running TrioExplore – Ensure path ‘c:\python24\
Scripts;’ in environment variables. And double click ‘start_te_server.bat’ under Trio-1.0\explorer.
◦ At the command line, you are now prompted for an admin user login to PostgreSQL, which should have been created along with your PostgreSQL installation and which will be used by TrioExplorer to create new user roles and database instances.
◦ TrioExplorer should now be reachable from your browser using http://localhost:8080/. For new users can now press ‘Create a new user’ and create their own Trio login and database instances, which are then managed by the PostgreSQL server.
Configuration of Trio System Configuration of Trio System (cont)(cont)TrioPlus
◦ Create new PostgreSQL user role and database instance Run ‘createuser demo’ Run ‘createdb demo’, the name must be the same as
username
◦ Initialize Trio schema information for new user by access as same as windows superuser authentication to access PostgreSQL. Use TrioExplorer will be easily just press ‘Create new role’ in Web.
◦ Connect to new Trio database using the command line clients by running ‘python trioplus.py –u demo –d demo –p’
Trio API Integrated Into Other Trio API Integrated Into Other Python ScriptsPython ScriptsTrioCnx
◦ TrioCnx(pgdb) –PyGreSQL connection pgdb
◦ cursor() – return a new TrioCursor object for the current connection.
◦ commint() – commits the current transaction.
◦ rollback() – performs a rollback for the current transaction.
◦ close() – closes the Trio connection
TrioCursor◦ execute(triql) - Executes a TriQL statement triql for this
cursor object.
◦ fetchone() - Fetches a single XTuple object from the current cursor position.
◦ fetchall() - Fetches and returns a list of all XTuple objects beginning from the current cursor position.
Trio API Integrated Into Other Trio API Integrated Into Other Python Scripts (cont)Python Scripts (cont)XTuple
◦ len() - Returns the number of Alternative objects contained in this XTuple object.
◦ getAlternative(idx) - Returns the Alternative object at the designated index idx.
◦ getConfidence() - Returns the confidence value (if any) of this XTuple object as the sum of its Alternative objects' confidence values.
◦ getQuestionMark() - Returns whether this XTuple object has a question mark or not.
Alternative getLineage() – returns a list of immediate lineage
information traceLineage() – performs a transitive lineage traversal for
this alternative back to the base data getConfidence() - Returns the confidence value of this
alternative. computeConfidence() - Computes the confidence value
Trio API And Translator Trio API And Translator (Python)(Python)
TriQL built-in functions TriQL keywordsAggregate(), Alternative(), As(), Avg(), BinaryExpression(), Brackets(), Cascade(), ColumnList(), Column(), Command(), ComputeConfidences(), Conf(), ConfUniform(), ConfScaled(), Count(), CreateIndex(), CreateTableAs(), CreateTable(), CreateTempTableAs(), CreateTempTable(), DataType(), Dot(), DropIndex(), DropTableList(), DropTable(), Eavg(), Ecount(), EcountStar(), Emax(), Emin(), Empty(), Esum(), FromClause(), GroupAlts(), GroupbyClause(), GroupByKey(), HavingClause(), HorizontalSelect(), IdentifierList(), InsertList(), Insert(), InsertSpec(), Lineage(), Literal(), Lsum(), Max(), Maybe(), Maybe(), Min(), OrderBy(), Question(), SelectClause(), Select(), SelectOptions(), SetOperator(), Star(), Sum(), UnaryExpression(), UncertainSet(), UniformTable(), ViewTable(), WhereClause(), WithConfidences()
'all', 'and', 'any', 'as', 'avg', 'by', 'cascade', ‘compute', 'conf', 'confidences', 'count', 'create', 'distinct', 'drop', 'eavg', 'ecount', 'emax', 'emin', 'esum', 'except', 'exists', ‘flatten', 'float', ‘float4', 'float8', 'from', 'group', 'groupalts', 'having','in', 'index', 'insert', 'int', 'int4', 'int8', 'intersect', 'into', 'is', 'like', 'lineage', #'lavg', #'lcount', #'lmax', #'lmin', #'lsum', 'max', 'maybe', 'min', 'noconf', 'nolineage', 'nomaybe', 'not', 'null', 'on', 'or', 'order', 'scaled', 'scaledbyexp', 'select', 'sum', 'table', 'temporary', 'trio', 'trio_aid', 'trio_xid', 'uncertain', 'uniform', 'union', 'merged', 'values', 'varchar', 'view', 'where', 'with'
DemoDemo
TrioExplorer and TrioPlusTrio Query Language (TriQL) Structure
◦ Drop index/table◦ Create Trio table/index◦ TriQL language◦ TriQL manual:
http://infolab.stanford.edu/~widom/triql.html ◦ Sample: TriQL Script
Current Supported for TriQLCurrent Supported for TriQLSupported DDL and Insert Commands Supported Subset of TriQL Query Language Create Table Create Trio Table Create Index Drop Table - Including “Cascade” option to drop
derived tables• Drop Index• Insert Into T Values <data>
o With or without alternatives, confidence values, and ?
o T must be a base table• Insert Into T <TriQL query>
o T must be a base table
Select-Project-Join queries• Create Table T as <query> Select Distinct• Union, Intersect, or Except of two subqueries (duplicate-
eliminating)• Union All of two subqueries• Merged, Flatten, and GroupAlts Horizontal subqueries in Where clause
o Including shortcuts and aggregation/group-by/having• Horizontal subqueries in Select clause
o Including shortcuts and aggregation but not group-by/having
Conf() function with any number of tables or Conf(*) o except for Conf(*) in the SELECT clause in conjunction
with DISTINCT or MERGED Maybe() predicate “Uniform <table-name>” in From clause• “as conf” for query-defined result confidence values
o Including “Uniform as conf” and “Scaled as conf”• “Compute Confidences” at end of query• Lineage() predicate
o Not in horizontal subquerieso “=⇒” abbreviation allowed
• NoLineage, NoConf, and NoMaybe
Table of TriQL contentsTable of TriQL contentsULDBs Uncertainty attributes, maybe annotations and
confidence valuesSQL over ULDBs Selection, projection, join, subqueries, duplicate-
elimination, grouping and aggregation, aggregate variants, set operators, order by
Flatten and GroupAlts Flatten is used to turn tuples with alternative values into regular tuples, while GroupAlts is used to create or restructure alternative values
Horizontal subqeries: The [ ] construct [ ] in the where clause, [ ] with joins, Syntactic shortcuts in [ ], [ ] in the select clause, [ ] with Self-Joins
Built-in Functions Conf() and Maybe() Multi-table conf()
Result confidences Result confidence evaluation, uniform and scaled result confidences, On-Demand confidence computation
Built-in Predicate Lineage() The Lineage() predicate lets queries filter joined tuples based on whether they are related via lineage
Options Nolineage, Noconf, and NoMaybe Indicate lineage, confidence values, and/or?'s should be omitted from query results
Data modification Insert statement, delete statement, update statement
Advantage and DisadvantageAdvantage and Disadvantage
Advantage◦ Open source and free support for any non-benefit
users to experience new Trio DBMS◦ Advanced components in relational DBMS◦ Computing confidences◦ Efficient, Convenient, safe, Multi-User storage of
and access to, Massive, Persistent
Disadvantage◦ Time cost for query◦ Dependency◦ On development stage
Advantage and DisadvantageAdvantage and DisadvantageDisadvantage
◦ Time cost for query
Using ‘SELECT attr-list FROM X1, X2, ..., Xn WHERE predicate’ as a query example in [6] for a comparison between relational database and ULDB.
Over standard relational database: For each tuple in cross-product of X1, X2, ..., Xn
Evaluate the predicate If true, project attr-list to create result tuple
Over ULDB: For each tuple in cross-product of X1, X2, ..., Xn
Create “super tuple” T from all combinations of alternatives
Evaluate predicate on each alternative in T ; keep only the true ones
Project attr-list on each alternative to create result tuple
Details: ‘?’, lineage, confidences
Documenting Trio system Documenting Trio system report/bugsreport/bugs The install instruction in website
http://dbpubs.stanford.edu:8011/doku.php/trio:installation , indicated unclearly the version of Graphviz for Trio system. In Graphviz website http://www.graphviz.org/ only has version 2.16 but not compatable except version 2.14. Graphviz version 2.14 is available for download in http://infolab.stanford.edu/trio/code/graphviz-2.14.1.exe .
The windows authentication supperuser needs to be created first in the PostgreSQL in order to connect to database. After established the connection, TrioExplorer and TrioPlus can use the supperuser’s login and password as windows authentiction to access to database system. However, it doesn’t mention at all in the installation procedure on how to create this typle of new user. The only way to solve it is to use PostgreSQL->pgadmin III manually.
After created the supperuser, I have to modify some codes in setup.py in Trio-1.0->setup directory in order to run ‘python setup.py –u myname –d myname –p’.
TriQL query statements in http://infolab.stanford.edu/~widom/triql.html#options, there are many samples queries not working properly as desired.
ReferencesReferences M. Mutsuzaki, M. Theobald, A. de Keijzer, J. Widom, P.
Agrawal, O. Benjelloun, A. Das Sarma, R. Murthy, and T. Sugihara. Trio-One: Layering Uncertainty and Lineage on a Conventional DBMS. Proceedings of the Third Biennial Conference on Innovative Data Systems Research (CIDR '07), Pacific Grove, California, January 2007. Demonstration description.
O. Benjelloun, A. Das Sarma, C. Hayworth, and J. Widom. An Introduction to ULDBs and the Trio System. IEEE Data Engineering Bulletin, Special Issue on Probabilistic Databases, 29(1):5-16, March 2006.
Trio: A System for integrated Management of Data, Uncertainty, and Lineage. Retrieved on November, 18, 2007 from http://infolab.stanford.edu/trio/ .
PostgreSQL. Retrieved on November, 20, 2007 from http://www.postgresql.org/.
Python. Retrieved on November, 15, 2007 from http://www.python.org/.
Trio: A System for Data, Uncertainty, and Lineage. given by Jennifer at various venues, 2006-07. Ppt.
Questions?Questions?