emuleomo@yes.rutgers.edu (Emuleomo) (06/16/89)
Attn RTingres GURUs. I am an Informix kind of guy but lately I have been forced? to do some work in RTingres but I have run into a few _ROADBLOCKS_. a) I can't seem to be able to change the schema of a table once it is created. For instance; I need to increase the size of a name fld from c15 to c20. Also, if this can be done, will Ingres automatically rebuild the table like Informix will do? b) I am experimenting will QBF and VIFRED. Can you specify multi screen forms in VIFRED and use it in QBF? Your help is muchO appreciated!! Respondez S'il vouz plait at Email ==> emuleomo@yes.rutgers.edu --Emuleomo O.O. ** Research is what I'm doing when I dont know what I'm doing! ** -- ** Research is what I'm doing when I dont know what I'm doing! **
davek@rtech.rtech.com (Dave Kellogg) (06/19/89)
In article <...> emuleomo@yes.rutgers.edu (Emuleomo) writes: > >a) I can't seem to be able to change the schema of a table once it is created. > For instance; I need to increase the size of a name fld from c15 to c20. > Also, if this can be done, will Ingres automatically rebuild the table > like Informix will do? > Once a table is created its data definition is fixed. The way to change the size of an attribute is to use the copy command to copy out the table, drop the table, recreate it, and then copy it back in. I know of a few people who have written simple front-ends to do this for you in a forms-based fashion (it's not terribly hard), and perhaps there is a net-lander who has such a program to post. There is a performance argument for this approach because it allows the data to remain clustered (i.e. adjacent attributes stored adjacently) in at least one sense of the word. With many table re-orgs, it's essentially "pay me now" (drop/recreate) or "pay me later" (with pointers to attributes stored in different locations from the base table, increasing I/O). >b) I am experimenting will QBF and VIFRED. Can you specify multi screen forms > in VIFRED and use it in QBF? Yes. See the section in the QBF manuals on "joindefs" where you create (or specify) the joins between relations and then QBF will create a default [multi-table] form that you can edit with VIFRED. (I am assuming you mean multi-table when you say multi-screen). Dave Kellogg Relational Technology New York City 212-952-1400
bg0l+@andrew.cmu.edu (Bruce E. Golightly) (06/19/89)
Sounds like you're looking for an "alter table" command. It isn't there at present. We've kicked around the idea of writing a utility package to do this, but we don't have any manpower to spare for the task at present. I'm not sure what your second question means. If you want the user to be walked through a series of screens, you probably should be working with ABF rather than QBF. Note that the latest and greatest versions include user defined "pop-up" forms, which will allow the creation of really impressive applications. I'm going to add this feature to the next incarnation of one of my sub-systems to provide pop-up reference tables. Bruce
emuleomo@yes.rutgers.edu (Emuleomo) (06/23/89)
Just as a quick followup on my previous question <again for RTI gurus>
IS THERE A WAY TO IMPLEMENT a SERIAL fld in Ingres (as is done in Informix)??
A serial fld is a fld which has ALL the following properties.
(AS defined by Informix software)
a) It uniquely identifies a row in a table. i.e. it is a key for that row.
b) It is managed by the DBMS AUTOMATICALLY and a unique value is
generated for any new row inserted into that table.
c) It cannot be modified, once generated!
** I dont understand why I (a self proclaimed Informix Guru) am having _such_
a hard time with Ingres. Maybe I should stop thinking in Informix!
However, I thought that SQL is SQL is SQL! and RDBMS is RDBMS is RDBMS.
Respondez S'Il vouz plait.
--Emuleomo O.O. (emuleomo@yes.rutgers.edu)
--
** Research is what I'm doing when I dont know what I'm doing! **llojd@rivm.UUCP (J.W. Diesel) (06/27/89)
In article <Jun.23.12.01.07.1989.1388@yes.rutgers.edu>, emuleomo@yes.rutgers.edu (Emuleomo) writes: > > Just as a quick followup on my previous question <again for RTI gurus> > > IS THERE A WAY TO IMPLEMENT a SERIAL fld in Ingres (as is done in Informix)?? > > A serial fld is a fld which has ALL the following properties. > (AS defined by Informix software) > a) It uniquely identifies a row in a table. i.e. it is a key for that row. > b) It is managed by the DBMS AUTOMATICALLY and a unique value is > generated for any new row inserted into that table. > c) It cannot be modified, once generated! > > ** I dont understand why I (a self proclaimed Informix Guru) am having _such_ > a hard time with Ingres. Maybe I should stop thinking in Informix! > However, I thought that SQL is SQL is SQL! and RDBMS is RDBMS is RDBMS. > > Respondez S'Il vouz plait. > > --Emuleomo O.O. (emuleomo@yes.rutgers.edu) > -- > ** Research is what I'm doing when I dont know what I'm doing! ** As I would not want other INFORMIX gurus :=) to be mislead I cannot resist the temptation of reacting upon Emuleomo's interpretation of the definition of the informix serial datatype. a) To my experience the serial datatype DOES NOT enforce unique values (only a unique index will do this): on insertion of a row containing a serial column you may, optionally, enter an 'initial' value. This value will be used wether or not it value is unique !!! b) DBMS management means that, unless an 'initial value' is given, the inserted value is the maximum-value incremented by one. Informix seems to 'remember' the maximum value somewhere since deletion of row(s) containing the maximum value doesnot change the value to be assigned to future 'serial' insertions. c) For those who want to update a serial column: alter the datatype to integer, perform your updates and alter the column back to serial. I'm not sure if the 'remembered' maximum-value is affected by this action. Though I'm not familiar with INGRES-SQL you might consider the following scheme to insert unique values: INSERT INTO yourtable (serial_column) SELECT MAX(serial_column) + 1 FROM yourtable; Informix does not permit you to use the insert-table in the from clause, maybe INGRES does. You may of course use a temporary table to store the maxvalue but that would make you pay a considerable performance penalty *and* it can cause consistency problems: process A determines max; process B determines max; process A inserts serial (=max + 1); process B inserts serial ==> A and B insert same serial !!! -------------------------- Jan Diesel National Institute for Public Health and Environment Laboratory for Air Research mcvax!rivm!llojd -------------------------- Wonder how he got there with that name !!!!
moiram@tekcae.CAX.TEK.COM (Moira Mallison) (06/30/89)
In article <1389@rivm05.UUCP> llojd@rivm.UUCP (J.W. Diesel) writes: >scheme to insert unique values: > INSERT INTO yourtable (serial_column) > SELECT MAX(serial_column) + 1 FROM yourtable; > >Informix does not permit you to use the insert-table in the from clause, maybe >INGRES does. You may of course use a temporary table to store the maxvalue >but that would make you pay a considerable performance penalty... If the table has a heap or hash storage structure, the use of the max function will result in a sequential scan, which REALLY imposes a performance penalty on a table of any size. (This may also be true of the isam and btree structures, depending on how smart the query optimizer is. Dave?). I solved this problem by storing the next surrogate key value in the table with an artificial primary key value. ("DUMMY" in the name field in my case. I retrieve it with a direct access, increment it and update it. Moira Mallison Tektronix, Inc.