Download Database Interfaces under the Version 6 Engine Architecture

Database Interfaces under the Version 6 Engine Architecture '
Emily P. Wallace, SAS Institute Inc., Cary, NC
_ut
J=-or the last seyeral SUGI conferences, you have been hearing
the Multiple Vendor Arcllitecture (MVAl and the multiple
EH'tg:ine conoopts of V-arslon 6 of the SA$8 Systam that make·data
to work.with a small portion of the data so you need a way to subset it. Another significant problem is the security of the data and
the- impact on the parfQrm.a~ce for other users of the database
system;
access transparent. Thus far. we have implemented four databas, engines and developmeot Is underway on other cbItaba$a
engines, so you CBn see that the engine architecture has become
In. order to solve these problems, we have created the ACCESS
procedure. Its job is to capture all of the Information needed to
aOQEJSS a database. This information can inctude such things as
the name of the da1abase, the- -password and the other options
needed to uniq....1y Identify the database. II also allows you to
specify- the elements of the database you want to work. with and
give them SAS names that will be meaningful to you. You can
,pacify a where clause- to subset the records retrieved so that
you work with only the portion of the datal:Jase that you -need,
a reality. These new intarfaces provide much more functionality
and fl.axibil~ty than their counterparts 1n Version 5 of the SAS Sys-
tem.
The VerSion 5 Interfaces are generalty procadure-6tyle Interfaces
with each database hav~ its own set of procedures that bring
the da.la from tf\a Database Management System (DBMS} into a
SAS data set. You can run the SAS procedures against this copy
of the DBMS data to perform your analyses Q( produce your
reports. If there are procedures thai update the database, they
generally use SAS data sets as input for the updatas. Both the
Before we ·Iook at PRQC ACCESS in more: detail, we need to
define two new terms. master descriptor and subc1esGrJptor. A
master descriptor is a member of a SAS data library that contains
a layout of the database and all DBMS information needed to
access Its data. A master descriptor can point to bOth tables and
views In relational database systems. A $ubdescnptor or view is
based on a master descriptor and contains a list of elements In
the 'database thaI will be used by the SAS program. the SAS
namas and formats associated with the elements, and any subsetting or ordering information that Is valid for the DBMS. The
subdescriptor or view is tlsed by thoa procedures Of DATA. S1ep
to access tile DBMS data.
eKtract and updata pr'Qcedures are very flexible rn allQwing YOI,.I
to work will1 a subset of your data. However, the on~y way to keep
the SAS data you have extracted from your database C(Jrrent Is
to perkxllcally .extract the DBMS data. For many people that is·
fine, but there are others who want to access- the DBMS data
directly from SAS programs without mBklng a -copy -of the data.
This Is one of the primary reasons. fOr the develOpment of the mul-
tiple engine architecture.
The primary goal of the multiple engine architecture is to provide
transparent access to data that are not stored in SAS data sets,
What this means is that you can use the same SAS programming
You must use PROC ACCESS to CrBat!;! mastElr descriptors and
subdescrlptors._lf you wantto k&epthe descriptors you buila, you
must Stlpply a lIBNAME statement pOitltlng to a SAS data library
.where you want the. descnpiOf'S to be wriUen.lfyou do not supply
a UBNAME statment. the descriptors will 'be written to the WORK
library and will be temporary. To invoke the procadure, type
statements on-data regard.less of where the data Bre stored. With
the muttlple englne- architecture, you can execute DATA and
PROC steps directiy against the DBMS data- instead of executing
against a copy of the data. Now you use the same tool for analy~
sis, graphkls, and reporting needs. A long term benefit of this new
architecture is that as the SAS System adds new fea1l.R"es and
addltional·capabllities, these new features will be immediately
avaUable to the DBMS engine users as welt
-;
PROC J.CCE5s. i
~n the program edit window· and submit the statement. You see
the ACCESS panel that displays tile cootents of all SAS data
libraries you have defined to your SAS sesslon. Master descriptors are Identified by their member type (M.mtype) of ACCESS
and subdescrlptors are. Identified by their type of VIEW. Views
oreatod by the SOL procedure are also Idenmied by the type of
VIEW and will also be displayed on the ACCESS PANEL
The database engine. makes. the direct access possible. The job
of the database engine is 10 accept any request from 1he ·SAS
$upervisor, translate it into. rtle appropriate DBMS n:lq1,l9st, and
return data Or status information to ttte SAS 5uparviSOl' in the correct SAS format. For example; W a procedore asks th<!SAS
supefVlsor to read a reoord, the database engine wm build a
DBMS read request, pass it to the database system, and receive
the data that was read. Then the engine must translate the DBMS
data into the format that the SAS System-expects and pass the
translated data to the SAS supervisor wtIo passes ~t on to the procedure.
Tl'IiS description of a database. engine migh1 make you belteve
thet it Is a relatively simple task to bring DBMS daia into the SAS
environment but that is not entiraty true. There are sotne p~ob­
lems that must be handled to make the process work smoothl~.
on.a of the most visible problems. is the difference ln the rutes .for
SA.S names and the rules for DBMS names. '" many cases, the
DBMS names are longer than the SAS names and cootaln ·characters that 'Would -not be valid in SAS variable names. Also, there
is no syntax on the procedure statements to specify- the name
of the database that oontaios the data you want-to use. Many
DBMSs aloo have optiOnS that must be set by the calling pro.gram
before accessmg ttle databa" management system. In many
cases, the databases con1ain large amounts of data and you want
347
d;e.=d~
I
I
Libll_...
~1PI
I ~ WVSD Pll)I'nI a.~
II _=~x
g~=~~
;Uu.o:m l«Il£
: =~
I
_ I
:=n~:
IIlIHIPI ftllll"
l
' . .Ion "IIIII'
,l, ,,
To create a subdescrlptor from the master descriptor you just
cr~atedt type a C beside 1tte name of the master deSQiptQr on
the ACCESS PANEL. You see the SUBOESCRIPTOR OISPLAY
PANEL. whi(lh lists all of the- database elements. tha.t are eligjble
for inclusion in10 the lJiew. It the master desaiptor has assigned
&AS names for th-e elemerrts in the database:. tI""J(J SAS names are
displayed and cannot be Changed. '(0 select elements. enter an
S in the FUNC field 10' each el"",ent you want. If the SAS name
field Is blank. you may fill in a name to usa or you can let the pr0cedure assign the ~am&S.
PMlL _________________________________________ t
: C_!III •• ->
h~lI
I
:
II
II
.
,,,
,,
:
,,
I,
+-_______________________________________________
,i
+-------- ------_._-_._--_._-
,I C_1I4 ••.,
11 ____ .-f
1
~Am
i "tI1l1OUurftlll 10: 1t.III101ll 1ULI Ili\llJj tRIP-II
IIiUIt, .s;ll<ltRIP HR, VIII!
iI LIJI!.'dI1,:I
DllfPQl'SAII DUA IItl LIIIIMII - - . . . . . , - I
roc m CCWIIII IMI
WIWII
roaw
I
Cij$!DIID....J:tUD.
n.
iI
-.::~
-",.
-n
..
~ft~.
Rl'!'mllUi'AH
I\ft...IIIIU 1Wl'I1.
Tho ACCESS PANEL of the ACCESS I'roooduro
t
To create a: new master descriptor. type the foUowing on the command line of the ACCESS PANEL:
,
I
1114 5IIBH5'1lmR DlIPW
I
sc_ 1
---~---------------------+
811f1
AlKtIML1'IaftJ:I
I
,
UI~; DI.2
I
I
:
I
I;VftCUI
I
~ SIS.
:
j
;
I
,
1
:
\
1
t-.-----------~---_~
---------~---------:l-----+
CREATE Iibnalrn3'.memfflU'I1e.ACCESS
where memname is the member you want to create and IIbname
points 10 the Dbrary where you want the mom"'" stored. II you
have more than one database engine installed on YOUI' CPU, yOLl
wiil see a list of the engines you have and you should choose the
one needed-for the master descriptor YOU_Bfe Cr'ElElting. If you
. have only ono database ongine installed, you wm bypass this
panel. l"ha nan step is to identify the name of the database or
table 10 be dascrlbed by this master descriptor. Once this info,·
rnfiltiOfi Is supplied, you sea the MASTER DESCRIPTOR
DISPLAY PANEL, which lists the names of the elements of the
database. If you want to ellforce Itie use of standard names for
al.sulXlescriptors created .rom this master descrlptDr, type In the
$AS names you want m be used. If you <10 not supply SAS
namas, then subdascriptors can use SAS names their choosIng. If you do not want an alement to be ayailable for use in any
·views, type an N In the selection f~td and the element Will be
marked as non-display and·cannot participate in anyviews. When
you ootar ttta END I)Q(J"Jn'I&nd. the new master descriptOf Ie created and·can be used to create subdeSCfiptors.
Screen 3
The SUBOESCRIPTOR OISPLAY PANEL Is used
to Specify the Columns fo!' the VIeW
If you want to define 5ubsetting or ordeling information, type
WHERE on the command line. The SELECTION CRITERIA
ENTRY PANEL is displayed and you can type In where clauses
or ordering clauses that are valid for the DBMS you are acces~
Ing. When you exit the SUBOESCRIPTOR OISPLAY PANEL, the
new subdescriptor is writt&n and its name Is dls~ay.ed 00 theACCESS panel. Now youI' new view is ready for' use by SAS pro-cedures and the OAT A step. II you want m validate the view
belore loavlng the ACCESS PANEl., you can invoke the FSEDIT,
FSBROWSE. or FSVIEW prooedul'es to ·see the data from the
DBMS.
o.
+11LIt1'11lll alfllll:i" Un, PMn----------------------------------------_+
,I
o
~lJf(!l ~~t
.wra1U,,.10l ID1 IIWln
I 11DH1: J:
~
,I
I
II
I ~HtluU{lll - 'IWI ru.lt:IlCO Q ' aM:
:
Qplrt-.rLt.l.t ... utftll! '1919.0'_11' llId 'I!S'_I._U'
:
_lin IW'mI
I
~:
-nn.
1:5111: PJl2
J'OUl1'
- ...
__
-1121_
"'
, ............
..........
..
IIU1W'l'lfIII
:DAft'.
_ _ WifU •
.."......,..
,,
AIIlUJ:LTIClIf.JIO
_ _ '1S.
,
I
:
I
t-_.
_________
I
:
I--·--------_~_~~~
_____ ___________________·___
~
2
_______________________________________________
1 ______ I,
,
i:------..I
ac......
_
r
,
,I,
,,,
,,
,,
,,
,,
:
j
lCCW
iUDIII
11:1. C01.IJA!WI!
I
,
PlDl
Wld lWIIl DlH
'r:1!I'S
I
i~d~.~
..l~-nd.
------------------------------------------_._---------.
..
The MASTEI'I OESCRIPTION OISPLAY PANEL
. Ilsts the columns in the DB2 Table
4
The SELECTION CRITERIA ENTRY PANEL
Specifies the Rows to be Viewed
Now that your view is bultt, you can use it In SAS programming
statements that will acceS$ data in your database. The program-
348
to PROC ACCESS that wm let Y"" dafine the file layouts 50 that
we can use that information to build the master descriptors. This
feature wlll also enable you to create new databases using a data-
mlng statements are the sarna statements that you would use to
""""ss BAS data set.; tha only difference Is that your DATAoption points to a subdeocriptor Instead of a physical data library.
If your database descriptors are In a permanent SAS data set,
you must use a LlBNAME statement to point to U1e library. If they
are In a _porary library, no UBNAME statement is required.
base layout you define. For example, you might want to create
a new ORACLE· table using the NEW= opjlon In PROC FSEOIT.
Curtenuy this is not Sl,Ipported. but in the future you will be able
to dafine a new ORACLE table using PROC ACCESS to build a
master descriptor and than create the table usIng PROC FSEOIT.
LIIIAIIE 1 'WIPIf.DE9-CRIP.LIBltARY';
noc PRlft ElA~A.X.SUGI'rRI:P:
llUll';
Another enhancement planned for a future release Is the ability
to build a database view containing Information from more than
one database of the sama DBMS type whenttte databa!i1e system
can do the joins efficiently. Since PROC SOl already cr&ates
views that join two of our database views. we will let PROC SOL
do the jofns of database views whoean the database cannot do the
iolns itse~.
PROC ACCESS and the <Ie"",ipto'" an> pow.orful tools boo,us.
they allOw you to work with !he DBMS data without a detaiIeCJ
knowladgo of database concepts and terminology. Once tho
descriptors are set up, they remain valid until the database struc~
tuno """"ges. Thensfore, thare Is very Iittio maintenance required.
The descriplOfS can be used as a security mechanism to control
both the elements and the recol'ds that you can see. The master
descriptors and the vtews can be kept in separate data fibraries
so the database admCnistraticn group can control the building of
masts< doseriptors and ""bdescnptors and provida access to
only the infonnatlon to th8 databases that you need.
The database el'lQinea that we currently have
lmp~mented are
for DB2~ under MVS, SQl./DS~ under CMS, ORACLE and
Rdb{VMS· undar VMS· . The.. produ"'" are all port of tha
SAS/ACCESS- family of interface products and they will be available with Rol . .se 6.06 of the SAS Systam.
There are probably some of Y"u who .J<> 'oV1>f>d..-Ing why we
changed the way Wf3 InterfaQel to databases. For B yar~ of rw..
sons, poopl. preferred to extract da.a from the DBMS Into a SAS
data .ot and work with the copy oftha data. In Version 6 you con
still do the ext.racte. and iii Soma casas this approach is better
than doing direct access from procedlJr&s and the DATA step.
If you are going to use the same data in many procedures In ttle
same SAS .session. itwould probably be advantageous to extract
the data-into a SAS data set. SAS data seta are organized In a
manner that provides optimum performance for SAS proooduros
and the DATA step so procedures using SAS data set input will
frequently take laBS CPU time than procedures that use views to
access DBMS infonnation.
Fu1ure engines will Inetude interfaces to at! currentiy supported
database systems (IMS, DLfl, SYSTEM 2000C' Data Managemef1t
Software, IDMS/R~ DATACOM/DBO and ADABASO) plus other
databases you request. You will also see the existing engines
ported 1:0 other platforms when there Is adequate interest in them,
AS you carl see, the mtdtiple engina architecture provides you
with a powerful and flexible Interface to your database $ystem.
You can write SAS programs to access your DBMS data without
knowing specific DBMS tenninology or DBMS programming conoopts. YOu use the same programming statements wheth8( your
data are in SAS data sets or database systems. Also you can
choose between difect access and extracts based on your pro-
gramming taSkS.
Another situation where you mlght prefer to do an extract of ths
DBMS data is when your database is large and you want 10 read
• large portion of the rocords. If the database, being shared by
other users on an on"na system, your program might have a negative lmpact on the system performance and cause the other
s.o.S, SYSTEM 2000 and SAS/ACCESS are registersd trademarks of SAS InstitLJte Inc., Cary, NC, USA
users to experience stow response time. 'The Database AdmlnlsImtiOn group might suggest that Y"u dO extra"'" of the DBMS
data into SAS data sets tor security reasons also. You shOUld
carefully ovaluate the types of BAS programs Y"u will use to
ac::oess the DBMS data and doo1de whirther' to use dil"5t;:t aecess
ORACLE is a registered trademark of Oracle Corporation.
or database extracts based on the programming tasks and your
company's guIdelines.
DATACOM/DB Is a registered trademark of Computer
ales International.
PROC ACCESS solves many problems for SAS users but the",
.... some situtatiOns that ~ still does not handle. If your data are
in • database or fife system that does not conlaln daoc<lptivo
InfonnlltiQo as a p:art of the database sy$tem , sl,ICh as IMS,
VSAM, tape foles. or _
extemol files, you might want to use
PROC ACCESS and an onglno to work with your dala. One of
the thll1gs we wRl be wortdng on ill tile future is an erdlQflcement
ADABAS Is • registered tradamork 01 Software AG..
RdbfVMS and VMS are .rademart<s of Digital Equipment c0rporation.
Associ~
DB2 and SOLIDS are trademarks of International Business
Moohln.. Corporation.
10M SIR is a regislOred trademark of Cullinet Software, Inc.
349