1. Why a tool like this¶
Has this ever happened to you?
- Resubmitting to another journal, you resized ten figures and their text, one by one.
- After reanalysing the data, the text said "1,240" but a figure legend still said "1,212".
- A reviewer asked "Which data and criteria is Figure 3c based on?", and six months later you could not remember.
- A co-author asked for "bigger text, please"; once you changed it, that figure no longer matched the others.
The key idea: fig-atlas treats a figure not as something you touch up by hand but as a result rebuilt from settings. Each of the scenes above then ends with "build again".
What the tool promises¶
| Promise | What changes |
|---|---|
| Drawn from settings | Colours, sizes, axis ranges and layout are all in a settings file (the config). To change a figure you do not open the figure or the program: you change one line of the config and build again. What you changed stays in writing, so it is easy to undo. |
| Drawn at printed size | A figure is not drawn large on screen and then shrunk; it is drawn 1:1 at its size on the page (the A4 text width). Write 7 pt in the config and it is 7 pt on paper. When the journal changes, you change one page setting. |
| Checked automatically | Every build looks for overlapping or clipped text, text that is too small, colours that colour-blind readers cannot tell apart, and font formats journals reject. |
| Its sources are recorded | Each figure records which data files (down to a value unique to each file, like a fingerprint), which settings and which software made it. When data change, you are told which figures are stale. |
| Numbers come from the data | Counts, percentages and statistics in a legend are not typed by hand; they are filled in from the analysis. When the data change, the legend changes with them. |
| It follows the paper's course | Exploratory figure → promoted to a paper Figure → figures no longer used are archived. Why the storyline changed, and the data behind each change, are recorded too. |
What the tool does not do¶
- It does not choose analysis methods or statistical criteria. Which test to use and which threshold is right are yours to decide; you write them in the config.
- It does not decide the paper's story. Which figure becomes Figure 1 and what you claim are your judgement. The tool records those decisions and checks that the figures match them.
- It does not change raw data. Raw data are only read. Making them look better (shorter names, other units) is done in the config when the figure is drawn.
So what does a researcher need to learn?¶
Not much. By the end of this guide you will be able to:
- prepare data as a bundle and read its check results (chapter 3)
- make a figure from a bundle, from code you wrote, or from a sketch of the figure you want (chapters 4 and 5)
- read and change the config (chapter 6)
- read the check results and decide what to fix
- promote a figure to a paper Figure and keep a record
Previous: 0. How to read this guide · Next: 2. Overview