On the the missing diagrams in Category Theory (first-person version)

Abstract

Most texts on Category Theory are written in a very terse style, in which people pretend a) that all concepts are visualizable, and b) that the readers can reconstruct the diagrams that the authors had in mind based on only the most essential cues. As an outsider I spent years believing that the techniques for drawing diagrams were part of the oral culture of the field, and that the insiders could read texts on CT reconstructing the "missing diagrams" in them line by line and paragraph by paragraph, and drawing for each page of text a page of diagrams with all the diagrams that the authors had omitted. My belief was wrong: there are lots of conventions for drawing diagrams scattered through the literature, but that unified diagrammatic language did not exist. In this chapter I will show an attempt to reconstruct that (imaginary) language for missing diagrams: we will see an extensible diagrammatic language, called DL, that follows the conventions of the diagrams in the literature of CT whenever possible and that seems to be adequate for drawing "missing diagrams" for Category Theory. Our examples include the "missing diagrams" for adjunctions, for the Yoneda Lemma, for Kan extensions, and for geometric morphisms, and how to formalize them in Agda.