• How to have a software product without a dedicated datacenter. Every developer could run the entire system on their desktop in Docker (with a subset of obfuscated production data), and when pushed to production “it just works”
• Super-short release times. It was not unusual that I’d finish coding a feature after lunch and it would be released before the end of the day
• Incredibly frequent releases. There were some days where we would release over 20 features to production. Coming from organizations that struggled to define and implement bi-weekly releases, this was a revelation and amazing to see first-hand a real-world example of releases as described in The Phoenix Project

Each of these is a big lesson, but probably the biggest lesson is that the “Principal Software Engineer” role has wide variations, not only between organizations but sometimes within the same company.

If your organization is looking for a problem solver with deep experience in software engineering using Java, Python, and SQL please keep me in mind.

Because the users needed the fix as soon as possible, I reviewed with the head of the desk and responsible development manager the half-hour option, as well as a more expedient option.

Since we could contain the risk by making a copy of the executable, and because it was used only by this trading desk, the expedient solution was chosen.

Emacs is well known to safely binary files because it doesn’t add characters (e.g. trailing newline, word-wrapping, etc.), I brought up the executable in Emacs; the fix only required substituting a smaller string for the existing one. I searched for the string in the file (C-x C-f) and made the change. nul characters were added at the end of the new string, making sure that I had the exact same number of characters including the original trailing nul.

After making sure that the file was the same size, I crossed my fingers while saving the file (C-x C-s) over the original one and had one of the traders restart the application. (Other traders already running the app weren’t affected because the original version was still at the same inode.)

The app started fine and the bug was fixed, so the other traders restarted theirs and had no problems either. Because I fixed this problem in about 5 minutes I was hero for a day and became known as a go-to guy for problems.

(original as bullet points 3/10/17)

I’ve been using Oracle DBMS for years, and sometimes it’s hard to differentiate what is an “Oracle built-in” and a standard SQL feature.

In teaching a Database Modeling course at RIT in MySQL I have found out that there are some differences between Oracle and MySQL:

• MySQL comparisons are case-insensitive, so SELECT * FROM table WHERE name='ABC'; returns the same results as SELECT * FROM table WHERE name='abc';

• MySQL table names are case-sensitive but only on Linux.

• MySQL has a richer set of datatypes, especially for integer types: tinyint, smallint, mediumint, int, and bigint, but is missing BLOB/CLOB and native(-ish) XML.

• Single and double-quotes are allowed for string literals (but cannot be mixed).

• Executing queries in an Oracle database are done by specifying the database name in the connect string. MySQL supports the USE command to choose a database to run.

• There are a lot fewer functions available than in Oracle (which someone could write a an-oracle-function-a-day series and have a year’s worth of material).

• CHECK constraints are “supported” in DDL but before MySQL 8 are only verified to be syntatically correct.

• MINUS is not supported, but can be simulated with NOT IN (<subquery>)

• Concatenation operator || isn’t supported, must use CONCAT(). MySQL allows string literals to be concatenated without operators (SELECT 'foo' 'bar';).

• tables can be aliased by backticks with or without the AS keyword because “ and ‘ are both allowed for string literals. Oracle lets you use a "quoted" value for a table alias without AS, for example SELECT 'foo' AS foo_col, 'bar' "bar_col" FROM dual;

• nested queries must have a table alias in MySQL even if table name isn’t referenced. For instance, AS my_sales is mandatory in this statement: SELECT count(*) from (SELECT distinct * from sales) AS my_sales;

• Joins inside nested queries need to have unique column names in MySQL. Oracle will rename the columns (generally by appending “_1” to the duplicate column name) but an error is returned. For instance, given tables STOCKTRADES(cusip, price, trade_date, commission) and BONDTRADES(cusip, price, trade_date, commission). In Oracle, SELECT count(*) FROM (SELECT * FROM stocktrades UNION SELECT * FROM bondtrades) AS all_trades;

• NVL() is not supported, use COALESCE() instead

• Good summary is at this SlideShare but it’s sometimes not clear what the original Oracle analog would be and it’s a little dated.

And the code:

<script src="https://unpkg.com/mermaid@8.0.0/dist/mermaid.min.js"></script>

<div class="mermaid">
graph TD;
    A-->B;
    A-->C;
    B-->D;
    C-->D;
</div>

Adapted from http://kkpattern.github.io/2015/05/15/Embed-Chart-in-Jekyll.html

As a Systems Programmer you need to know several things;

How to do something, arguably the easiest to learn.
Why do I do it.
When do I do it.
When don’t I do it.
When to follow the rules.
When not to follow the rules.
How to make the process foolproof.
How to make the process damn foolproof, or as close to it as possible.
How to think inside the box.
How to think outside the box.
Knowing which one to use.
How to get out of any mess you created.
How to get out of any mess someone else created.
When to ask for permission.
When to ask for forgiveness.

It’s as much as an art as it is a science. We can teach the science, the art of Systems Programming has to be lived.

(From a quote by Mark Jacobs on LinkedIn posted to rands-leadership.slack.com, I can’t find original post)

I don’t like the default github themes because they’re missing basic things like sidebars and displaying post titles. These can be added (because it’s all code nothing is impossible), but it requires modifying a theme’s _layout and locking you into a theme, but theme flexibility is the point of using a built-in theme.

Github has recently-ish (11/2017) added shared themes a.k.a remote themes, allowing specially-built themes stored in github to be used by reference in your _config.yaml.

I chose Minimal Mistakes because it looked good and didn’t need any customization in _layouts (none in my case) and minimal configuration.

Some of the steps I needed to do to incorporate it:

• Changes to _config.yml
  • added required new plugins
  • defaults for posts
  • Added sidebar links and avatar that links to my Gravatar page
  • Declarations for pagination, breadcrumbs,
  • Include for _pages dir to get it processed
• created _pages/category-archive.md and _pages/tags-archive.md to generate pages with tags and categories
• removed layout: post from front matter in every post (now defaulted in _config.yml)
• Added a lot of defaults for posts and pages in _config.yml
• Removed all customizations from _layouts
• Moved about.md to _pages and fixed the permalink in the file
• Created _data/navigation.yml to set up a menubar
• Changed index.html to have “important” front matter, the key being layout: home. Until I added this the homepage didn’t work.
• Add a header image and configure header code in index.html front matter

It looks like a lot of steps, but was a good learning experience for Jekyll and a lot of these steps were because I was porting from another theme.

I will be labeling the merge to master branch so I can easily find the changes, but for now all changes are made on the branch minimal-mistakes.

Following the instructions here will enable one to choose an R kernel. Note that there is a ton of output from R when running install.packages("deftools") because it’s building it from source.

If one is not doing a lot of R development then create an environment that can be switched into using source activate <env-name> when R notebooks are needed. I do this because R is a ton of code that could crap up your root Python install.

(This is an alternative description of how to set up R in Jupyter and includes steps that work for Jupyter Notebooks and Jupyterlab.)

I’m going to be moving content that’s been in draft form for “a while” (a long while in some cases), I’ll mark them the date I published them as the as-of date. If I’m disciplined enough I’ll add the original date in the text.

At some point I’m going to go on a theme hunt again. I have to add an ‘about’ page with my resume and contact info, and find a template that shows the title and date at the top of a post. I’ve added them in myself to an old verison of the website, but then you can’t auto-update anymore (or at least I haven’t figured out how to do it yet).

(Via: Medium.com/writingcooperative.com)