@@ -259,13 +259,30 @@ Reader options
259259 require different kinds of images. Currently this option only affects
260260 the markdown and LaTeX readers.
261261
262−`--filter=`*PATH*
262+`--filter=`*EXECUTABLE*
263263: Specify an executable to be used as a filter transforming the
264264 Pandoc AST after the input is parsed and before the output is
265265 written. The executable should read JSON from stdin and write
266266 JSON to stdout. The JSON must be formatted like pandoc's own
267− JSON input and output. Filters may be most easily created in Haskell,
268− using the utility function `toJsonFilter` from `Text.Pandoc`.
267+ JSON input and output. The name of the output format will be
268+ passed to the filter as the first argument. Hence,
269+
270+ pandoc --filter ./caps.py -t latex
271+
272+ is equivalent to
273+
274+ pandoc -t json | ./caps.py latex | pandoc -f json -t latex
275+
276+ The latter form may be useful for debugging filters.
277+
278+ Filters may be written in any language. `Text.Pandoc.JSON`
279+ exports `toJSONFilter` to facilitate writing filters in Haskell.
280+ Those who would prefer to write filters in pandoc can use the
281+ module `pandoc.py`: see <http://github.com/jgm/pandoc-filters-python>
282+ for the module and several examples. Note that the *EXECUTABLE*
283+ will be sought in the user's `PATH`, and not in the working directory,
284+ if no directory is provided. If you want to run a script in the
285+ working directory, preface the filename with `./`.
269286
270287`--normalize`
271288: Normalize the document after reading: merge adjacent