2020-04-25 05:00:08 +00:00
|
|
|
<!doctype html>
|
|
|
|
|
<html>
|
|
|
|
|
<head>
|
|
|
|
|
<title>Zk | 2015-04-21-done-vs-finished</title>
|
2023-05-11 05:35:08 +00:00
|
|
|
<link rel="stylesheet" href="https://fonts.googleapis.com/css2?family=Gentium+Plus&family=Lato&family=Ubuntu+Monodisplay=swap" />
|
2024-05-05 04:25:08 +00:00
|
|
|
<link rel="stylesheet" type="text/css" href="/style.css?2024-05-04" />
|
2020-04-25 05:00:08 +00:00
|
|
|
<meta name="viewport" content="width=device-width" />
|
|
|
|
|
<meta charset="utf-8" />
|
|
|
|
|
</head>
|
|
|
|
|
<body>
|
|
|
|
|
<main>
|
|
|
|
|
<header>
|
|
|
|
|
<h1>Zk | 2015-04-21-done-vs-finished</h1>
|
|
|
|
|
</header>
|
|
|
|
|
<article class="content">
|
|
|
|
|
<hr />
|
|
|
|
|
<p>type: post
|
|
|
|
|
date: 2015-04-21
|
|
|
|
|
slug: done-vs-finished
|
|
|
|
|
title: Done vs. Finished - On Code Reviews and Learning Languages</p>
|
|
|
|
|
<hr />
|
2020-06-24 07:10:07 +00:00
|
|
|
<p>As part of my yearly goals at work, I’ve been doing more and more work in Go.
|
|
|
|
|
It’s a pretty fantastic language, I’ve found, and fairly easy to pick up, minus
|
2020-04-25 05:00:08 +00:00
|
|
|
a few little caveats (goroutines took me a bit to wrap my head around). A lot
|
|
|
|
|
of this has been taking place in a little project,
|
|
|
|
|
<a href="https://github.com/juju/jujusvg">jujusvg</a>, which takes a Juju bundle and
|
|
|
|
|
outputs an SVG representation of that bundle for displaying in webpages such as
|
|
|
|
|
<a href="https://jujucharms.com">jujucharms.com</a>. This was a project I started on a
|
|
|
|
|
sprint and have been shepherding through its growth over the last several months
|
|
|
|
|
with the help of several members of our team who are more well-versed in Go than
|
|
|
|
|
I am.</p>
|
|
|
|
|
<p>In particular, a recent branch of mine has brought up a lot of interesting stuff
|
|
|
|
|
to think about, especially when it comes to learning a new language and what it
|
|
|
|
|
means to be finished.</p>
|
|
|
|
|
<p>The goal with <a href="https://github.com/juju/jujusvg/pull/23">#23</a> was to allow
|
|
|
|
|
embedding of charm icons in the bundle SVG so that the diagram was totally self
|
|
|
|
|
contained, not requiring any requests to outside assets such as the icon SVGs
|
2020-06-24 07:10:07 +00:00
|
|
|
through SVG’s <code><image></code> tag. The main benefit to this is that anyone can build
|
2020-04-25 05:00:08 +00:00
|
|
|
a bundle diagram and embed it anywhere - a slideshow, a webpage, whatever -
|
|
|
|
|
where it will work without network access. In addition to this, we can now do
|
|
|
|
|
some checking on whether or not an icon exists, is really an SVG, and so on. In
|
|
|
|
|
the case of the charm store, we can also embed icons by fetching them directly
|
|
|
|
|
from GridFS, rather than over HTTP, which will be a significant speed boost.</p>
|
|
|
|
|
<p>Within our projects on the Juju UI Engineering team (and, I suspect, elsewhere
|
|
|
|
|
in the company), we utilize a system of reviews required to land code. In
|
|
|
|
|
particular, we require at least one QA, and at least two positive reviews in
|
|
|
|
|
order to land any branch. If there are any negative reviews, they must be
|
|
|
|
|
addressed and received positively before the branch can land.</p>
|
|
|
|
|
<p>In the case of #23, the process of reviewing stretched out over 21 days and 124
|
|
|
|
|
comments, which is exceptionally long for a branch to live in review (we
|
|
|
|
|
generally aim to have a branch completed - from start to reviewed to landed -
|
|
|
|
|
within one day), but in this case, I think that the reviews helped me to learn
|
|
|
|
|
how to be a better Go programmer, and a better programmer after all.</p>
|
2020-06-24 07:10:07 +00:00
|
|
|
<p>This, I’m sorry to say, felt slightly unusual for me. I think that the industry
|
|
|
|
|
of software development is strongly imbued with “guess culture”. In this
|
|
|
|
|
dichotomy, social interaction between actors is divided into either “guess
|
|
|
|
|
culture” or “ask culture”, which describes the ways in which the actor will act
|
2020-04-25 05:00:08 +00:00
|
|
|
when confronted with a problem they do not know how to solve. In software,
|
2020-06-24 07:10:07 +00:00
|
|
|
you’re expected to either a) know how to solve the problem, or b) know how to
|
2020-04-25 05:00:08 +00:00
|
|
|
figure it out yourself.</p>
|
|
|
|
|
<p>This was shown to me early on in an interview question:</p>
|
|
|
|
|
<blockquote>
|
|
|
|
|
<p>Given a private jet in a hanger, find out how much it weighs.</p>
|
|
|
|
|
</blockquote>
|
2020-06-24 07:10:07 +00:00
|
|
|
<p>Answers ranged from “cut it in half again and again into smaller parts until you
|
|
|
|
|
can place one on a scale” as a sort of Zeno’s Paradox of plane-weighing, to
|
|
|
|
|
“fill it with a known amount of jet fuel and fly it until it crashes.”</p>
|
|
|
|
|
<p>The answer that the interviewer was looking for, however, was “Google it”.</p>
|
2020-04-25 05:00:08 +00:00
|
|
|
<p>I think this is an interesting question when it comes to the idea of ask versus
|
2020-06-24 07:10:07 +00:00
|
|
|
guess culture. On the one hand, it points to the obvious solution of “sure,
|
|
|
|
|
it’s good to be able to solve a problem given limited resources, but don’t be
|
|
|
|
|
afraid to just ask,” but it also highlights, in a simple-minded sort of way,
|
2020-04-25 05:00:08 +00:00
|
|
|
that you should be able to sort the problem out by yourself without opening
|
|
|
|
|
yourself up to criticism by asking what might be a stupid question.</p>
|
2020-06-24 07:10:07 +00:00
|
|
|
<p>Counter to guess culture, though, is the concept of ask culture, wherein it’s
|
|
|
|
|
acceptable to ask for help, clarification, or instruction. There’s a lot bound
|
2020-04-25 05:00:08 +00:00
|
|
|
up with this, though, and the ideas of machismo so prevalent in software
|
|
|
|
|
culture, as well as the idea that we should leave university with 100% of what
|
|
|
|
|
we need to know in order to be successful, both of which are ludicrous when
|
|
|
|
|
investigated with the least amount of scrutiny.</p>
|
|
|
|
|
<p>While I scoffed at it early on, I think that Agile methodologies (we follow a
|
|
|
|
|
Lean Kanban style) do a good job of subtly encouraging a shift from guess to ask
|
|
|
|
|
culture within a team at the very least, and sometimes within a whole company.
|
2020-06-24 07:10:07 +00:00
|
|
|
The idea behind Agile isn’t specifically to provide upper management with
|
2020-04-25 05:00:08 +00:00
|
|
|
burn-down charts and instant results, but to create an environment where
|
|
|
|
|
knowledge-sharing, rather than knowledge-hoarding, is the norm. A lot of the
|
|
|
|
|
other aspects of Agile fall out from that: cross-training, quick turn-around,
|
|
|
|
|
daily stand-ups, and so on.</p>
|
|
|
|
|
<p>In the case of reviews, the requirement to get two thumbs up and address any
|
|
|
|
|
concerns means that a conversation is automatically required between developers.
|
|
|
|
|
I think that worked really well for us on this branch, not just because I got to
|
|
|
|
|
be a better Go developer, but also because others on the team know exactly how
|
|
|
|
|
to implement my proposed API changes, and get to understand just how the library
|
|
|
|
|
itself works on a lower level than they might if they just looked at the
|
|
|
|
|
example.go file.</p>
|
|
|
|
|
<p>The only place we found ourselves stumbling happened to be purely mechanical:
|
2020-06-24 07:10:07 +00:00
|
|
|
GitHub, while it’s great in a lot of respects, did not lend itself well to this
|
2020-04-25 05:00:08 +00:00
|
|
|
review turn-around. Other tools, such as Rietveld, allow both inline comments
|
|
|
|
|
as well as inline responses, and unified messages back and forth. With GitHub,
|
2020-06-24 07:10:07 +00:00
|
|
|
I found myself having to reply to each comment left on my branch with “done” in
|
2020-04-25 05:00:08 +00:00
|
|
|
order to track which I had responded to in code, which meant that all of my
|
2020-06-24 07:10:07 +00:00
|
|
|
coworkers got a list of emails containing simply “done” and a link to the pull
|
2020-04-25 05:00:08 +00:00
|
|
|
request. Better, I think, to have a single email sent with my replies
|
2020-06-24 07:10:07 +00:00
|
|
|
addressing each comment once I pushed my subsequent branch. Again, it’s simply
|
2020-04-25 05:00:08 +00:00
|
|
|
a mechanical thing, and not that tough to deal with, but I feel as though there
|
2020-06-24 07:10:07 +00:00
|
|
|
are other code review tools out there that work better than GitHub’s.</p>
|
2020-04-25 05:00:08 +00:00
|
|
|
<p>What it really came down to, after all of the reviews, was that I needed to be
|
2020-06-24 07:10:07 +00:00
|
|
|
more than just productive in Go, I needed to be fluent. There’s a lot that can
|
2020-04-25 05:00:08 +00:00
|
|
|
be said either way when it comes to coding styles, but there are some objective
|
|
|
|
|
benefits to language fluency. In particular, my approach to fetching icons took
|
|
|
|
|
the following path</p>
|
|
|
|
|
<ol>
|
|
|
|
|
<li>Fetch synchronously, one icon at a time, over HTTP in all cases.</li>
|
|
|
|
|
<li>Implement <code>IconFetcher</code>s for retrieving icons how the user specifies,
|
|
|
|
|
defaulting to HTTP</li>
|
|
|
|
|
<li>Allow fetching icons concurrently using goroutines, using one set of channels
|
|
|
|
|
for synchronization and another for errors.</li>
|
|
|
|
|
<li>Allow fetching icons concurrently using goroutines and
|
|
|
|
|
<a href="http://golang.org/pkg/sync/#WaitGroup">WaitGroups</a> to simplify synchronization.</li>
|
|
|
|
|
<li>Since concurrency was either on or off, threads could get out of hand, so
|
|
|
|
|
allow fetching icons concurrently up to a limit using
|
|
|
|
|
<a href="https://godoc.org/github.com/juju/utils/parallel">github.com/juju/utils/parallel</a>,
|
|
|
|
|
which vastly simplifies the code overall path of the code, abstracting out a lot
|
2020-06-24 07:10:07 +00:00
|
|
|
of (but not all of - we’re not trying to be magical) the asynchronous patterns.</li>
|
2020-04-25 05:00:08 +00:00
|
|
|
</ol>
|
|
|
|
|
<p>Each of these iterations was shepherded along through suggestions and code
|
|
|
|
|
snippets from my reviewers. The point being was that I was writing Go code in
|
|
|
|
|
order to be productive while not yet being fluent. Through the help of others,
|
|
|
|
|
I was able to step all the closer to Go fluency so that I could make a branch
|
|
|
|
|
that did more than just get the job done by any means necessary, but also do
|
|
|
|
|
what I needed to in a clean, easy to read, and efficient manner, a perfect
|
|
|
|
|
example of utilizing an ask instead of a guess method. Which brings me to my
|
|
|
|
|
last point: done versus finished.</p>
|
|
|
|
|
<p>When I first proposed the branch, even though I proposed it as a WIP branch,
|
|
|
|
|
knowing that there would be comments, it was technically done. It fetched all
|
|
|
|
|
of the icons needed for a bundle, embedded them in the SVG diagram, and produced
|
2020-06-24 07:10:07 +00:00
|
|
|
a valid output. However, it wasn’t <em>finished</em>.</p>
|
2020-04-25 05:00:08 +00:00
|
|
|
<p>If we had wanted or needed to iterate on this branch faster than we eventually
|
2020-06-24 07:10:07 +00:00
|
|
|
did, then I’m sure we could’ve landed it as-is on each go-around with a
|
2020-04-25 05:00:08 +00:00
|
|
|
follow-up card created to represent future work. My activity graph on my
|
2020-06-24 07:10:07 +00:00
|
|
|
profile might’ve looked a little more full, but I think that having something
|
|
|
|
|
finished was much more important, for a library such as this. We didn’t want to
|
|
|
|
|
leave things hanging too long and risk releasing something which wasn’t
|
2020-04-25 05:00:08 +00:00
|
|
|
complete, even if it did what it said on the tin.</p>
|
|
|
|
|
<p>The responses that I got on my code review were basically all about that. It
|
|
|
|
|
was a little painful to have a branch lingering for 21 days, but on the other
|
2020-06-24 07:10:07 +00:00
|
|
|
hand, it’s much more satisfying to land something that I’m proud of than
|
2020-04-25 05:00:08 +00:00
|
|
|
something I feel more like sweeping under the rug. Through the help of Roger
|
|
|
|
|
and Martin, I was able to do that. Not only was the branch more finished than
|
|
|
|
|
it was simply done, I became a better developer in the process, learning more
|
|
|
|
|
about what is the Go-ish thing to do specifically as well as what is a good
|
|
|
|
|
CS-ish thing to do generally.</p>
|
|
|
|
|
</article>
|
|
|
|
|
<footer>
|
2024-05-05 04:25:08 +00:00
|
|
|
<p>Page generated on 2024-05-04</p>
|
2020-04-25 05:00:08 +00:00
|
|
|
</footer>
|
|
|
|
|
</main>
|
2020-06-24 07:10:07 +00:00
|
|
|
<script type="text/javascript">
|
|
|
|
|
document.querySelectorAll('.tag').forEach(tag => {
|
|
|
|
|
let text = tag.innerText;
|
|
|
|
|
tag.innerText = '';
|
|
|
|
|
tag.innerHTML = `<a href="/tags.html#${text}">${text}</a>`;
|
|
|
|
|
});
|
|
|
|
|
</script>
|
2020-04-25 05:00:08 +00:00
|
|
|
</body>
|
|
|
|
|
</html>
|
