Taking a sabbatical to work on Ruby profiling tools | Hacker News Reader

Taking a sabbatical to work on Ruby profiling tools

jvns.ca

128 pointsbkudria35 comments

Sabbaticals should be more condoned in the workforce. I suspect we may see less turnover in many areas of software engineering if this was an option after say a 3 or 4 year tenure at some company.

Absolutely agree. And if they are unpaid, and the employee can schedule them around projects finishing (or at least in places where multiple people are responsible for them so that the loss of one person isn't a big deal), there's really minimal cost to the company, but a potentially large upside.

May be even better if the unpaid sabbatical was a holiday, and paid ( or at least support ) if the period was used to contribute to open source?

If the unpaid sabbatical were a paid holiday it would be regular paid work?

I'm working on surveying existing Ruby & Python profiling tools so that I understand what's been tried already.

If anyone has tools/techniques I've missed, I'd love additions to this list:

https://gist.github.com/jvns/b81eb039f6373bc577d7dbdd978581b...

https://github.com/evanhempel/python-flamegraph is similar to pyflame but does its sampling at the Python level, that makes it somewhat costlier & less precise but it's not restricted to Linux.

I think sabbaticals are great. If you can, save up and do them yourself. Or, if you are switching jobs and have savings, ask for a month or two between the end date of your current job and the start date of the new job. It's a heck of a lot easier to spend money when you know you have a new job coming up.

What to do? Something technical is great. Non tech stuff is good too (travel, learning new skills, family time, volunteering, etc).

OT: HN folks, I posted this a few days ago, not sure if there was an issue with dup checking: https://news.ycombinator.com/item?id=15832094

On HN a story doesn't count as a dupe unless it has had significant attention. Until then, a small number of reposts is ok: https://news.ycombinator.com/newsfaq.html. This is one way we try to mitigate the randomness of what gets noticed.

>I’ll be on sabbatical from my job. . . and working on building better profiling tools for Ruby (and maybe Python??).

I can't comment on Ruby's profiling stack, but I think that cProfile is a great tool for finding bottlenecks in Python. Usually, I've found that sorting functions by cumulative execution time is sufficient. You can't do flame graphs out-of-the-box, however.

> I can't comment on Ruby's profiling stack, but I think that cProfile is a great tool for finding bottlenecks in Python.

Whereas I don't. Between the performance impact (code with and without cprofile can have very different… profile), the lack of stack information and yet the lack of precise, actionable code-level information makes it not really useful/convenient for non-trivial codebases. Feeding the trace files to more visual systems (e.g. gprof2dot or runsnakerun) doesn't help much either.

These days I tend to use python-flamegraph (or pyflame but that's linux-only, though it has the advantage of using ptrace and can be attached to a running process at any point), this gives a mile-high view of the entire software in an extremely readable format, and from there it's possible to dive into specific hotspots with limited-scope profiles (whether with cprofile, with line_profiler[0] or with dedicated benchmarks).

[0] which has cprofile's issue of impacting the software it observes except significantly worse)

General rule I have: sampling profilers for CPU performance, instrumenting profilers for memory and other uses, like analyzing control flow.

Sadly, Ruby has a great profiling tool already built in: dtrace integration. It's the Linux side that's lagging in this case.

dtrace is too low level to be really useful, have a look at Go and Java for powerful tools.

It does provide high-level (function, exception, etc.) probes as well: https://ruby-doc.org/core-2.3.3/doc/dtrace_probes_rdoc.html

Then you can either use existing scripts of build on top of that. It's quite powerful. What do you think is missing?

Compilers & runtimes can provide high-level static probes for dtrace or systemtap.

In fact CPtython 3.6 includes such probes: https://docs.python.org/3/howto/instrumentation.html

Good luck, mate. Looking forward to see some decent and helpful tools in the foreseeable future.

Oh man. I'm so excited about this. I love all of Julia's work. Would love to see the fruit of this sabbatical.

Good luck Julia! Hope to see some blog posts of your findings along the way.

> https://jvns.ca/blog/2016/06/12/a-weird-system-call-process-...

> does this tool, that can spit out Ruby stack traces quickly for any Ruby program with debugging symbols enabled, actually exist somewhere and I just don’t know about it?

To your question in your other post you linked in your post: Maybe you mean APM, it stands for application performance monitoring. Search for "Ruby APM". Maybe the can inspire you for your tool. Good luck

This “Sabbatical” trend in software companies is really toxic. It’s a justification that allows companies to burn people out without looking bad. A sustainable interesting career doesn’t need a break. It’s a subtle admission that you’re overworked and your work is monotonous, which are both signs of a poorly managed company.

Sabbaticals are reserved for tenured professors who have been working for a lifetime so they can achieve something non-academic. It should not become a euphemism for breaks that burnt out programmers who lack time management skills take so they can get their life together.

> Sabbaticals are reserved for tenured professors

Why do you think so? Sabbatical means a break. It's common for professors, but why prevent other people from taking it? I took a sabbatical from my uni to do some real work (and ended up moving unis, countries as a result), then took a sabbatical from work for personal reasons and did other things I wanted to do.

If that prevents a burn out for someone, great! Not every company has enough projects going on to reshuffle you to the place you enjoy. (Or may not want to for valid reasons) Not every sabbatical is about burnout either.

> Why do you think so?

Because... that's what they actually are. A sabbatical was a very specific thing that was hijacked by HR people in tech companies. What you did was take a break; that has very different connotations than what a sabbatical is supposed to be.

This may be a cultural thing, but I've always understood sabbatical as a long break - nothing explicitly related to academia. Multiple definitions I see on the internet seem to agree.

Well then I suggest you take a look at "The Origin and Early History of Sabbatical Leave"[1] -- its roots lie firmly in academia. It was re-appropriated (in probably the last ~10 years) by tech HR departments which are typically enamored with academia, so that's no surprise.

[1] https://www.jstor.org/stable/40222893

Let me rephrase: yes, the origins are in academia. What I meant is that the current definitions lost that bit. I think the meaning changed over time and often it's understood differently than in the 60s. I.e. currently it's not connected explicitly to academia, but rather a long term break from your current position in whatever sector.

I worked in an investment bank in the 1990s and they offered a sabbatical (and called it that) after 10 years and every 5 years thereafter IIRC. I think it was 2 months. At least 6 weeks. You had to take it all at once you could not subdivide it.

Companies shouldn’t burn people out in the first place. Calling this bandaid a “Sabbatical” is just marketing to avoid looking bad for their poor management. They should use another word and not tarnish the very old tradition of academic sabbaticals.

Again, why do you think that "academic sabbaticals" are any different? Academics work hard and get burned out, just like programmers. It's not a coincidence that professors take their first sabbatical not long after getting tenure.

A sabbatical is a break where you do something different with your life. If it's good for the mind of an academic, it's good for the mind of a programmer too. Having just taken a gap year myself, I recommend it to anyone. It should be more available.

Well, some people use the word specifically for academia and not white collars with white collar jobs and white collar burnouts and white collar lives thinking they are somehow worthy of being called a "scientist", an "architect", an "engineer" because they wrote a bunch of loops and felt tired and accomplished.

I couldn’t disagree more!

This sabbatical seems exactly like something design to allow an expert focus on something outside the general scope of work. To really dedicate oneself to an academic pursuit.

I fully support providing engineers a conduit to build something that helps everyone!

>Sabbaticals are reserved for tenured professors who have been working for a lifetime so they can achieve something non-academic.

For devs it could also be to work on something that isn't profitable.

Traditional "vacation time" in software companies is really toxic. It's a justification that allows companies to burn people out without looking bad. A sustainable interesting career doesn't need breaks scattered across the year. It's a subtle admission that workers are so burnt out they need a break every few weeks just to continue to function, which is a sign of a poorly managed company.

Contrast this with sabbaticals, which are given to tenured professors who have been working for a lifetime so they can achieve something non-academic. Why do we acknowledge that professors may want to achieve something outside of their workplace, while not giving programmers a sabbatical to do the same?

> A sustainable interesting career doesn’t need a break.

That's absurd.

I don't believe it's a trend. There were "sabbaticals" at companies I worked at in the 1990s. "Sabbatical" typically just means a longer-than-normal vacation that you don't get routinely and don't get until after you've worked at a company for a long time.

Reply on news.ycombinator.com