Skip to main content
Bower handles the code that sits alongside your experiments: scripts you upload, code Bower writes back to you, and code blocks inside your notes. This page covers what works, and one thing that does not.

​
Attach a script

Upload source files the way you upload anything else: drag them into a collection, use the upload button, or attach one to a Bower chat. The formats Bower accepts include .R, .Rmd, .Rnw, .py, .ipynb, .sql, .sh, .jl for Julia, .m for MATLAB, .do and .ado for Stata, plus .json, .yaml, .xml, .c, .cpp, .java, .js and .ts. If you tried this before and Bower turned you away, that was not you doing it wrong. Most operating systems have no registered file type for a .R or a .do, so your browser hands the upload over with the type field blank, and Bower used to refuse anything it could not identify. It now recognizes these files by their extension.

​
Read it without downloading it

Open the attachment and you see the file itself, syntax highlighted, with the language named at the top and buttons to copy or download it. An .Rmd keeps its chunk headers intact, options included, so include=FALSE is still visible where you wrote it. A few accepted formats have no highlighter yet, .css and .xml among them. Those still open and read as code, just without the colors. Files over 256 KB are not rendered in the browser: you get the file details and a download button instead, because syntax highlighting a file that size would lock up the tab.

​
Search inside your scripts

When a script is brought in through import, Bower reads the text inside it. Search for a function name you defined and the file comes back, and Bower can read the file when you ask about it, rather than asking you to paste the code into the chat.
This applies to files brought in through import, or attached in a Bower chat. A file attached directly to an existing note is stored and previewed, but its contents are not indexed, so it will not turn up in search.

​
Code in your notes

Code blocks in the editor are syntax highlighted, and each one has a language picker in its top corner. Pick R, Python, SQL or any other supported language and the block colors accordingly. Bower will not guess the language for you. A block with no language set renders as plain text, because a wrong guess quietly recolors your code and gives you no sign it happened. Short R snippets are the easiest thing in the world to mistake for something else. The language you pick travels with the note, so a block you tag as R is stored, searched and read back as R.

​
Code from Bower

Ask Bower for a script and you get it back properly fenced and highlighted, with copy and download on the block. Bower writes in the language of your work. If your files and notes are R, you get R back, rather than a Python translation of what you asked for.

​
What Bower can and cannot run

Bower can run Python, in a sandbox, for analysis over your data. Bower cannot run R. There is no R interpreter anywhere in Bower. Ask it to run an R script and it will say so, then offer the two things that actually exist: write the script out so you can run it yourself in RStudio, or do the equivalent analysis in Python and label it clearly as Python. It will not run something else and hand you the number as though it came from your code. This is deliberate. A result that came from a quiet translation of your analysis is worse than no result, because you have no way to tell it apart from the real thing.