systabinfo ti_flags and index partition

Greetings. Solaris 10, IDS 11.5 I am trying to write a utility to examine some internal structures within sysmaster. Specifically, a query concentrating on systabnames and systabinfo. I need to treat index partitions differently from a table partition. But how do I tell the difference? Well, according to information in sysmaster.sql, can find that in column ti_flags. Specifically: insert into flags_text values ('sysptnhdr', 262144,'Index Partition');That's 0x40000. The problem is my query seems to find no entries in systabinfo with this flag set: select * from systabinfo ti where bitval(ti.ti_flags, '0x40000') = 1 I get "No rows found", clearly an insane claim, since there are far more indexes in our server (or in *any* normal database) than there are tables. Oh, and I get the same "No rows" if I use the 262144. At the application level, I can tell an index fragment from a table fragment: The "indexname" column will have a name. But I had wanted to bypass the table fragments at the systabnames/systabinfo query to reduce the work at the application level. So my questions boil down to: 1. Why isn't this flag set on any partition in my server? 2. If I must accept this non-compliance with the sysmaster.sql setup, how else might I distinguish a table partition from an index partition within systabinfo? Naively yours (for believing the script) -- Jacob S.

Jacob, I believe a previous post discussed this issue: http://www.iiug.org/forums/ids/index.cgi/read/35580 Not pretty but you could rename all your indexes so that tabname in systabnames tells you if it's an index or not: example: ix1_customer, ix2_customer, ix3_customer, etc Steve > Greetings. > Solaris 10, IDS 11.5 > I am trying to write a utility to examine some internal structures > within sysmaster. >Specifically, a query concentrating on systabnames and systabinfo. >I need to treat index partitions differently from a table partition. >But how do I tell the difference? Well, according to information in >sysmaster.sql, >can find that in column ti_flags. Specifically: >insert into flags_text values ('sysptnhdr', 262144,'Index Partition');>That's 0x40000. >The problem is my query seems to find no entries in systabinfo >with this flag set: >select * from systabinfo ti >where bitval(ti.ti_flags, '0x40000') = 1>I get "No rows found", clearly an insane claim, since there are far more >indexes in our server (or in *any* normal database) than there are tables. >Oh, and I get the same "No rows" if I use the 262144. >At the application level, I can tell an index fragment from a table fragment: >The "indexname" column will have a name. But I had wanted to bypass the table >fragments at the systabnames/systabinfo query to reduce the work at the > application level. >So my questions boil down to: >1. Why isn't this flag set on any partition in my server? >2. If I must accept this non-compliance with the sysmaster.sql setup, >how else might I distinguish a table partition from an index partition > within systabinfo? >Naively yours (for believing the script) >-- Jacob S.

Here are a few hints: 1. If you look at syst= pnhdr where nkeys > 1 or nkeys=3D1 and npdata>1, then this partition = has both data and indexes. This was true from system upgraded from ea= rly version or system catalog tables. 2. If you look at sysptn= hdr where nkeys=3D0 then this is a data partition. 3. If= you look at sysptnhdr where nkeys=3D1 and npdata=3D0 then this is either a= n empty table or an index partition. John F. Miller III STSM,= Lead Architect [1]miller3= @us.ibm.com 503-747-1366 IBM Informix Dynamic Server (IDS)= [2]-----ids-bounces@iiug.org wrote: ----- = >To: [3]ids@iiug.or= g >From: "JACOB SALOMON" >Sent by: [4]ids-bounces@iiug.org >Date: = 12/29/2015 08:55AM >Subject: systabinfo ti=5Fflags and index partitio= n [36298] > >Greetings. > >Solaris 10, IDS 11.5 <= br>> >I am trying to write a utility to examine some internal stru= ctures >within >sysmaster. Specifically, a query concentrating= on systabnames and >systabinfo. >I need to treat index partit= ions differently from a table partition. >But how >do I tell t= he difference? Well, according to information in >sysmaster.sql, can = >find that in column ti=5Fflags. Specifically: >insert into f= lags=5Ftext values ('sysptnhdr', 262144,'Index >Partition'); >= That's 0x40000. > >The problem is my query seems to find no en= tries in systabinfo with >this flag >set: >select * fro= m systabinfo ti >where bitval(ti.ti=5Fflags, '0x40000') =3D 1 &g= t; >I get "No rows found", clearly an insane claim, since there are f= ar >more >indexes in our server (or in *any* normal database) = than there are >tables. Oh, >and I get the same "No rows" if I= use the 262144. > >At the application level, I can tell an in= dex fragment from a table >fragment: >The "indexname" column w= ill have a name. But I had wanted to bypass >the table >fragme= nts at the systabnames/systabinfo query to reduce the work at >the >application level. > >So my questions boil down to: = >1. Why isn't this flag set on any partition in my server? >2. If= I must accept this non-compliance with the sysmaster.sql setup, >how= else >might I distinguish a table partition from an index partition= within >systabinfo? > >Naively yours (for believing th= e script) > >-- Jacob S. > > >*************= ******************************************************** >********** = > Forum Note: Use "Reply" to post a response in the discussion foru= m. > > > References 1. 3D"mailto:miller3@us.ibm.com" 2. 3D"mailto:-----= 3. 3D"mailto:ids@iiug.org" 4. 3D"mailto:ids-bounces@iiug.org"