2018-04-19, Comments


TIMTOWTDI stands for “There is more than one way to do it”, an approach promoted by the Perl community.

The mindset behind it gets explored in more detail by the language’s creator, Larry Wall, in a talk given in 1999: “Perl, the first postmodern computer language”. He attributes the slogan to his daughter, Heidi, who says it’s a strategy which works well in her maths class; and she associates it with another saying used at school: “Tsall Good”. This doesn’t mean everything is good, or even everything has good bits. It means, overall, things are good. See the big picture.

Perl epitomises this. It’s eclectic and inclusive, supporting a variety of styles. One-liner? Fine! Like a shell script? Sure! Structured programming, object-oriented, functional? Why not! Tsall good.

I like that.

But do I feel that way about programming?


A contrasting mantra appears in the Zen of Python, a list of aphorisms which summarise the guiding principles behind Python’s design. Item number 13 states “There should be one — and preferably only one — obvious way to do it.”

Perhaps realising this sounds overly prescriptive, this rule is tempered by item 14, “Although that way may not be obvious at first unless you’re Dutch.”

Guido van Rossum, Python’s BDFL — Benevolent Dictator For Life — would be the Dutch person who finds things obvious. That’s right: Dictator. Programmers don’t like being told what to do any more than two year olds. How then has Python become so popular?

Maybe emphasis falls on should. There should be only one obvious way to do it: it’s just that — Dutch or otherwise — we haven’t got there yet.


For example, there is more than one Python. Obviously there’s Python 2 and Python 3, but it’s less obvious which to use. Don’t forget PyPy. Increasingly Python comes packaged with data processing and visualisation extensions, served up as a Jupyter notebook.


There is more than one program options module.

When I started with Python there was getopt, the one and only command line handler. Coming from a C/C++ background I was quite happy to use something resembling GNU’s getopt. Then optparse appeared. Now there’s argparse. All of these libraries are readily available. Which should I use? Not optparse, that’s deprecated, unless I’m already using it and it works, that is. Regarding the other contenders, the documentation archly notes:

Users who are unfamiliar with the C getopt() function or who would like to write less code and get better help and error messages should consider using the argparse module instead.

There are other non-standard Python options for parsing a command line too: ones which generate code from the usage notes, or by inspecting the code you want to expose.

There is more than one way to do it.


There is more than one unit test framework. The obvious one, unittest, like getopt, draws inspiration from elsewhere — in this case Java’s Junit. Unfortunately the port is too faithful, and you’ll have to inherit from super classes etc to test something. I much prefer PyTest, which flexes the language itself to deliver test assertions as asserts.

There’s also a doctest module in the standard library which executes and checks code found in strings (hold that thought!), and there are many other non-standard testing frameworks.

There is more than one way to do it.


There is more than one way of formatting strings.

As we’ve seen there’s more than one Python, and libraries are always up for reinvention. This is arguably evolution rather than a multiplicity of options. That is, the most recent way to do it should be preferred.

When it comes to string formatting, though, there has always been more than one way to do it, and more ways are still being added.

Do you use 'single' or "double" quotes for a string? """Triple""" quotes. Raw strings? Raw with an r or Raw with an R? TIMTOWTDI.

What if you want to embed the value of a variable in a string? Users familiar with C’s printf() function might prefer % formatting. Fans of $shell $parameter $expansion can use template strings.

Advanced string formattingstr.format — appeared in Python 3.0, backported to Python 2.6. No doubt it has advantages over % formatting, but for me it’s a little more obscure and a little less obvious. Python 3.6 introduces f-strings which build on str.format and knock down my reservations. The syntax allows you to evaluate expressions in strings: evidently Python is heading in Perl’s direction.


Let’s finish by returning to Perl, and to Larry Wall’s 1999 talk.

How many times have we heard the mantra that a program should do one thing and do it well?

Perl is not that program. Perl wants to do everything well. It integrates features and makes no attempt to homogenise them.

You’ve all heard the saying: If all you have is a hammer, everything starts to look like a nail.

Perl is no hammer: it has memorably been described as a Swiss army chainsaw, but Larry Wall likens it to a more conventional tool.

If all you have is duct tape, everything starts to look like a duct. Right. When’s the last time you used duct tape on a duct?

Python may aspire to offer a single obvious way to do something. It fails splendidly, being more duct tape than hammer.

I presented this blog post as a lightning talk at PyDiff a couple of days ago. The slides are here. The talk was recorded too: I appear about 24 minutes in.