Improving Wetware

Because technology is never the issue

The Web is no longer Open, DRM now part of W3C standards

Posted by Pete McBreen Tue, 19 Sep 2017 02:55:00 GMT

PostgreSQL version of ERD Query for GraphViz

Posted by Pete McBreen Tue, 01 Nov 2016 02:42:00 GMT

Finally got around to doing the PostgreSQL version of the ERD generator using GraphViz. This uses the database from the PostgreSQL Tutorial site

As before there is a PostgreSQLERDGeneration.sql query, the resulting DVDRental.gv dot file for graphviz to generate the file and the resulting DVDRental.html. Inline interactive version of the diagram is below.

A few differences

  • Primary keys are sorted to the top of the table symbols
  • Lines are thicker on hover to make it easier to select the relevant symbol
  • Query does not filter out empty tables.

This completes the set of databases that I have made this work for, might include DB2 at some point in the future if I ever work on an IBM system.

For this interactive version, hovering over the lines makes them larger so that you can click to highlight the line. This makes it easy to plan out a query by following the links between the relevant tables, regardless of where they are on the screen. A good example of this would be tracing out which language DVDs are rented out in a specified city? This needs seven tables and six relationships to determine this, and it is much easier to have the path highlighted while writing the query than having to remember the path as you write the query.

dvdrental actor_id first_name last_name last_update dvdrental.film_actor dvdrental.film_actor actor_id film_id actor_id film_id last_update>dvdrental.film_actor:film_actor_actor_id_fkey dvdrental.address dvdrental.address address_id city_id address address2 district postal_code phone last_update dvdrental.customer dvdrental.customer customer_id address_id store_id first_name last_name email activebool create_date last_update active dvdrental.address:address_pkey->dvdrental.customer:customer_address_id_fkey dvdrental.staff dvdrental.staff staff_id address_id first_name last_name email store_id active username password last_update picture dvdrental.address:address_pkey->dvdrental.staff:staff_address_id_fkey store_id address_id manager_staff_id last_update dvdrental.address:address_pkey-> dvdrental.category dvdrental.category category_id name last_update dvdrental.film_category dvdrental.film_category film_id category_id category_id film_id last_update dvdrental.category:category_pkey->dvdrental.film_category:film_category_category_id_fkey city_id country_id city last_update>dvdrental.address:fk_address_city country_id country last_update> dvdrental.payment dvdrental.payment payment_id customer_id rental_id staff_id amount payment_date dvdrental.customer:customer_pkey->dvdrental.payment:payment_customer_id_fkey dvdrental.rental dvdrental.rental rental_id customer_id inventory_id staff_id rental_date return_date last_update dvdrental.customer:customer_pkey->dvdrental.rental:rental_customer_id_fkey film_id language_id title description release_year rental_duration rental_rate length replacement_cost rating last_update special_features fulltext>dvdrental.film_actor:film_actor_film_id_fkey>dvdrental.film_category:film_category_film_id_fkey dvdrental.inventory dvdrental.inventory inventory_id film_id store_id last_update>dvdrental.inventory:inventory_film_id_fkey dvdrental.inventory:inventory_pkey->dvdrental.rental:rental_inventory_id_fkey dvdrental.language dvdrental.language language_id name last_update dvdrental.language:language_pkey-> dvdrental.rental:rental_pkey->dvdrental.payment:payment_rental_id_fkey dvdrental.staff:staff_pkey->dvdrental.payment:payment_staff_id_fkey dvdrental.staff:staff_pkey->dvdrental.rental:rental_staff_id_key dvdrental.staff:staff_pkey->

How to fail at Agile part 1

Posted by Pete McBreen Thu, 04 Aug 2016 02:45:00 GMT

In every iteration, have a few bugs that do not get fixed. After five or six iterations you can build up a reasonable size bug backlog without even trying, and the best bit is that you can hide them in the previous iterations so nobody important sees them.

Obvious fixes:

  • If there is anything left over in the current iteration, move it into the next and increase the priority of that item.
  • Review all items that overflow into the next iteration to make sure that the team understands what is needed.
  • Publish the failure up the management chain if a defect survives to iterations.

SqlServer version of Drawing ER Diagrams with GraphViz

Posted by Pete McBreen Sat, 19 Mar 2016 03:25:00 GMT

This is the same as the Oracle version, but works with SQL Server (tested on SqlServer 2012)


Only problem I’ve identified is that the databases that most need a generated ERD often are lacking in foreign keys that this query uses to identify the relationships…

Drawing ER Diagrams with GraphViz

Posted by Pete McBreen Sat, 19 Dec 2015 22:46:00 GMT

Recently as part of an archaeology task of understanding how some SQL queries were working, I needed to draw an ERD to help with my understanding of the database. After contemplating drawing the diagram by hand for a few seconds, I decided to leverage GraphViz and just draw a diagram of all of the foreign key relationships between the tables.

Since it was an Oracle database, the queries to read the relationships were not that complex ErdCrearion-specific.sql is designed to run in SQLDeveloper and prompt for the :OWNER tablespace name to pick the tables from, and limit the selection to the names mentioned in the tablelist CTE (unfortunately duplicated as I have not rewritten this to make it simpler).

From that query the output is pasted into the starter GraphViz .gv file scottsimple.gv

digraph ERD {
  graph [   rankdir = "LR" ];

  "SCOTT.DEPT" [ label="<SCOTT.DEPT> SCOTT.DEPT|<PK_DEPT>DEPTNO \l |DNAME \l LOC \l " shape = "record" ];
  "SCOTT.EMP" [ label="<SCOTT.EMP> SCOTT.EMP|<FK_DEPTNO>DEPTNO \l |EMPNO \l ENAME \l JOB \l MGR \l STARTDATE \l SAL \l COMM \l " shape = "record" ];

  "SCOTT.DEPT":"PK_DEPT"->"SCOTT.EMP":"FK_DEPTNO" [arrowhead = crow];

Which after being put through GraphViz

$ dot -Tsvg scottsimple.gv -o scottsimple.svg

gives us a nice image of the relationship. The table name is prefixed with the schema to make sure that you can identify the table correctly for those cases where the same table exists in multiple schemas, and the columns involved in the relationships are highlighted in their own box. The non-relationship columns appear at the bottom of the symbol (column ordering is maintained and hidden columns are not shown).


For a more interactive experience run the SVG file through this converter, svgconverter.rb which converts the SVG file to a HTML file with some D3JS JavaScript that highlights relationships and tables as you click on them - I found that this made it easier to trace through a model when there are 80+ tables.

$ ruby svgconverter.rb scottsimple.svg scottsimple.html

The resulting file when uploaded to a webserver that has d3.js in the right place is interactive - see scottsimple.html unlike the image above it can be clicked on to highlight the symbols or relationships - only the outer line of the table is clickable - the rest is left as an exercise for the reader.

Back to the future (again)

Posted by Pete McBreen Thu, 03 Dec 2015 04:27:00 GMT

Delving into the archives, Leslie Lamport wrote about the Future of Computing back in 2003. There are still too many programs that are not understood fully by their developers or maintainers.

Firefox is going to lose a lot of developers (and users)

Posted by Pete McBreen Thu, 26 Nov 2015 03:01:00 GMT

Somehow or other the Firefox community has convinced itself that scanning add-ons for vulnerabilities and malware is a good idea. Luckily Dan Stillman the developer of Zotero called them out on it pointing out that it is just Security Theater.

Firefox has always had lots of really large extensions, but by deciding that they must be signed and reviewed, the Firefox community has just committed itself to a LOT of extra work reviewing the extensions. Hence the dumb idea of scanning to see if there is anything malicious in it. Now that is an arms race that is going to be lost. The guys in the AdBlock game know that, a continual game of whack a mole. Actively developed extensions like Zotero really lose out because a manual review of a large codebase takes a long time, and scanning is insufficient (as the above link describes, it is easy to create an add-on that passes scanning and does nasty things).

Breaking news...programmers are not engineers

Posted by Pete McBreen Sat, 07 Nov 2015 23:59:00 GMT

The Atlantic just realized that programmers are not engineers

Sorry to break it to them, but this has been a topic of conversation long before I wrote the book Software Craftsmanship, which was published nearly 15 years ago.

I had forgotten about the Coding Horror website

Posted by Pete McBreen Thu, 30 Jul 2015 20:28:00 GMT

And then this article about testing — Doing Terrible Things To Your Code reminded me to look at it again.

QA Engineer walks into a bar. Orders a beer. Orders 0 beers. Orders 999999999 beers. Orders a lizard. Orders -1 beers. Orders a sfdeljknesv.

I sure wish more programmers would focus a lot of attention on testing their own code before passing it on to QA/Test. That way the QA/Test team can focus on finding the requirements and interaction defects, rather than the simple coding mistakes that are often the bane of their existence

Kevlin Henney - Seven Ineffective Coding Habits

Posted by Pete McBreen Mon, 16 Feb 2015 02:49:00 GMT

Kevlin Henney - of Curly Bracket Languages fame has a good video of his presentation at a recent NDC conference Seven Ineffective Coding Habits of Many Programmers. As usual a very entertaining talk, but Kevlin is also spot on in identifying ways in which we are lead to make incorrect decisions about the code we are writing.

In it he references a paper from Rob Pike Notes on Programming in C. Although Rob Pike wrote that paper back in 1989 it is still relevant, as can be seen by his words about variable names:

Length is not a virtue in a name; clarity of expression is.