I find this sentence confusing: it looks like it's expecting the dir to contain hidden files (starting with a .) for nodes and edges.

We should add a preliminary comment like, "given a graph specified as a couple of files called g.edges.csv.gz and g.nodes.csv.gz" (or whatever you like instead of g :-)).

While at it, we should describe the expected syntax and semantics of those two files. The syntax is a gzip-compressed, csv file, using space as separator and the following columns: bla bla bla. The semantics is partly pointing to SWH PIDs and adding to that the fact that the nodes file should contain a sorted list of all the (unique) node identifiers that occur in the edges file.

active voice :-), i.e. "To start graph compression:"

i'm guessing graph_name here means that the inputs are graph_name.{nodes,edges}.csv.gz, right?

if so, specifying the base file name as mentioned above would clarify this part too

We should briefly explain what these mapping files are, and state what their file names will be.