I’m using Tigergraph developer version on a Macbook Pro under Ubuntu Linux running in VMware Parallels Desktop Lite.
I have a favorite dataset of ASH dumps from a 4-node Oracle RAC database which I’ve used before for dataviz experiments and other purposes, so I thought I’d design a schema that could swallow a bunch of the data and I could run GSQL versions of queries I’ve done in standard SQL. The data are csv and I attached header lines to the files. They each have ~50K lines of data and about 30 data values per line.
After getting my schema designed and the data load to work and before diving into complex GSQL queries I added simple “COUNT()” queries against the edge and vertex types in my schema at the end of the load job. While doing this using cut-paste I forgot to change edge types on one of the pasted COUNT() queries, i.e. my script executed the same count query twice consecutively. While noticing this I also noticed something very alarming, which is that the same count query executed twice consecutively gave two different answers. Of course this cannot be the case so now I wonder whether there is a wrong results bug here or if I completely misunderstand something, which seems unlikely given the simplicity of the test case.
Please see the attached screenshot with 3 terminal windows visible: 1/ the two count() query results for the “in_WAIT” edge type, 2/ the two count() queries, and 3/ the two load lines where this edge type are loaded.
Re: 3 the fact that two LOAD lines are used to load a single edge type should not affect the count() but seemed like perhaps it could be relevant (?)
Thanks for any insights. Very impressed with how fast and easily I was able to get my schema designed and data loaded. The GSQL doc seems quite good.