David White
davey@autistici.org
davidnwhite.com/resources
To wrangle: Round up, herd, or take charge of.
Art of wrangling patterns in nature.
…complicated.
… a way to wrangle science.
Where could programming improve your science/ work more efficiently?
Programming is an interface to computers.
Hides some technical side of things, through abstraction.
low level - close to the hardware
high level - abstracted, less 'code technical'
Scientists? Sometimes all levels…
99.9% on a computer.
99.9% computer programming?
| Test | Experiments are designed and presented in Psychtoolbox |
| Measure | Subject data is saved in a comma separated file (CSV) |
| Pre-process | Subject data is verified, anonymised, and conglomerated |
| Analyze | ANOVA analysis computed using R builtin procedures |
| Draw | Figures are plotted using ggplot package |
| Publish | Manuscript is written and typeset using LaTeX. |
| Lab-notebook | Jupyter notebook |
Pushing towards fluency in several languages.
Not as scary as it sounds.
Very similar languages.
| MATLAB | Julia | R | Python |
|---|---|---|---|
| i = 3 | i = 3 | i <- 3 | i = 3 |
| v = [9, 7, 3] | v = [9, 7, 3] | v <-c(9, 7, 3) | v = np.aray([9 7 3]) |
| length(v) | length(v) | length(v) | len(v) |
Procedural, Object-Oriented
Operations
Identical!
What kind of interface do you want?
There is no ultimate language
Different languages for different purposes
Most languages can do anything you want,
but may not be the best choice for the problems
| Fortran | 1957 | Designed for scientific computing |
| Unix* | 1973 | Solidified conventions for personal computers |
| C | 1978 | Solidified conventions for programming |
| Matlab | 1984 | Designed for numeric computing |
| Perl | 1987 | Designed for general use by power-users |
| Bash Shell* | 1989 | |
| Python | 1991 | Designed for learning how to program |
| LAPACK* | 1992 | SciPy, Matlab |
| R | 1993 | Designed for statisticians |
| Julia | 2012 | Designed for scientific-computing |
Strong majorities with:
Julia catching up
Some people use
| C | Perl | python | Matlab | R | julia | |
|---|---|---|---|---|---|---|
| *Community | 2 | 3 | 3 | 1 | 2 | |
| *Data-tables | x | x | x | |||
| *Notebook | x | (x) | x | |||
| *Conventions | 2 | 3 | 3 | 1-3 | 3 | |
| *OOP support | 1 | 2 | 3 | 1-2 | 2 | |
| *Free (Freedom) | (x) | x | x | x | ||
| Free (Beer) | x | x | x | x | ||
| compiled | x | |||||
| interpreted | x | x | x | (x) | ||
| Linear Algebra | ||||||
| Tables | ||||||
| Package management | 2 | 3 | 3 | 1 | 3 | |
| Syntax simplicity | 1 | 1 | 2 | 3 | 1 | |
| Syntax efficiency | 1 | 3 | 2 | 2 | 2 | |
| Builtin-Editor | x | |||||
| Dynamic Interpreter | (x) | x | x | x | ||
| GPU | x | x | x | |||
| General Scripting | (x) | x | x | (x) | ||
| documentation | 2 | 3 | 3 | 1-3 | 2-3 | |
| C interface | 3 | ? | 2-3 | 2 | 3 | |
| Live Interpreter | ||||||
| Style |
1=worst
3 = best
Matlab fails some of the important ones
non-network local connections being added to access control list
[class=firefox]
i3-msg: Could not connect to i3 on socket /run/user/1000/i3/ipc-socket.2027376: No such file or directory
i3-msg: Could not connect to i3 on socket /run/user/1000/i3/ipc-socket.2027376: No such file or directory
https://hyperpolyglot.org/numerical-analysis
Very loose
Good - Freedom!
Bad - Freedom!
What can a language do out of the box?
What setup is required?
Library/package
A language may be technically inferior in all regards,
but may be the right choice based upon community
and available packages/libraries
Matlab - Psychtoolbox, sandbox
Python - Citation management
R - Regression
Stan - Advanced stats
Julia - Machine learning
C/C++ - Matlab + speed
eLisp - Text editor customization
Unix scripting (i.e. Bash, Zsh)(aka Terminal)
Perl - Text processing and database management
No secret.
Differences are semantic
If you learn one language in and out,
it is much easier to learn another