Understand every expression.
Be consistent.
Order everything like a story.
Trim continuously; when in doubt, delete.
Name variables with two characters and pluralize by suffixing with the number of preceding "_" + 1 "_"
-
"Thing":
th(pluralth_); ifthis taken, use the next character:ti(ti_). -
"Number of things":
ut. -
Dictionary (key => value):
ke_va(ke_va__). -
Dictionary (key => values):
ke_va_(ke_va__).
Uppercase global variables: GL (GL_).
Name functions explicitly; get_signal_to_noise_ratio(n1_, n2_).
Prefix names not to be exposed with "_".
When getting a file, keep its original name.
When writing a file
-
Name tables like
feature_x_sample_x_number.tsv. -
Name dictionary like
key_x_value.json. -
Add extension to mark a transformation like
feature_x_sample_x_number.log.tsv.
Import, at the top, only what is used in the file.
Count from one: 1, 2, ...
Use < and <= instead of > and >=.
Allocate less.
Use concrete, immutable data.
Use small data.
Initialize data structures ahead of time.
Compute less.
Be simple.
Be flat.
Minimize nesting like f1(f2(f3)).
Be predictable.
Do not change a variable's type.
Write small, pure functions, each doing one thing and returning one type.
Avoid accessing global variables within a function.
Store related data close in space.
Access related data close in time.
Handle errors close to their source.
Instead of creating a nested path, create intermediate directories sequentially.