I am on a contract in Oslo (commuting from Stavanger region), to document and help rewrite an older SQL Server system (lots of TSQL stored procedures).
System is poorly documented, currently extracting all possible job, job step, executable (stored procedures), table and column information into Neo4j, to find out exactly what is connected with what (documentation).
First intro'ed to Neo4j in a Stavanger contract, extracted Windows Active Directory information and directory information to model security rights for a digitalizing system (quickly finding all accessible documents on subjects, depending on user role, across servers, systems, and countries). Also extracted POSC/Caesar (ISO15923 data model for Oil and Gas) into Neo4j for demo purposes ... powerful stuff.
Found Neo4j to be fascinating.
that's a really interesting topic. Which tools did you use for the extraction? Something generic or specific SQL queries? I always wanted to add that kind of analysis to jqassistant.org
You should connect with Joe Chesack if you don't know him yet, he's been using Neo4j for a long time in Norway and in your field too.
Hei Michael -
Yeah, Joe was the main Neo4j dude in the digitalization project.
Thanks Karin! <3 <3 <3
Charles, if you are still sometimes in Stavanger, check us out, every second tuesday of the month, 5:30-7:30.
And I've been thinking about your egg steamer puzzle, that it takes less water to steam more eggs. We can model this in Neo4j, and you can present it at the meetup.
RSVP here for the next meetup.
How have you been getting on with your documentation project? It is an excellent use case. But I would say that because I am looking at it from the other end. You are dealing with the physical and I am trying to group logical processes and entity types (if you will forgive my relational terminology). Have you used any similarity or community detection algorithms on your content? If so, any interesting findings?
Hei Douglas -
I have loaded the node and relationships of everything of the following structure - the property represent documentation and are at wildly various states of completion (Some Executable nodes lacking any description or comment, for example).
Excutables are either Job or Stored Procedure, and DatabaseObject can be table, view, or column.
Altogether some 3700 nodes and relationships, I think.
I've populated the data using Talend to extract from SQL Server queries then insert into Neo4j, extracts of comments and descriptions from source code, and manual definitions of Jobs lying outside SQL Server (Talend and manual jobs).
I tried using Tom Sawyer software to produce visual representations of job hierarchies, but the product was too expensive for our project.
I am now looking at GrandStack to produce a "Job Website" for both documentation and CRUD of the Neo4j base. A bit of a struggle, as I am new to GraphQL and React. Trying to rewrite a sample project, but it's been a struggle.
Not a fnut of "similatiry or oommunity detection algorithms".
Hope you're not disappointed.
You might also want to take a look at the NASA Knowledge Graph use case.
@david.meza01 wrote this pretty in-depth blog post a little over a year and a half ago. He's very much a thought leader for these types of applications.
@Christophe_Willemsen did a Online Meetup that might be interesting to you.
And NodusLabs created a tool to check out text relevance.
How can I be disappointed? Unravelling the chaos with which you seem to be confronted is hard enough but then trying to transpose the information into a set of tools/technologies/ with which, I suspect, you are unfamiliar is great. I wish you well. I hope you find yourself in a position to show-off a little bit to everybody. I will read it!