The beginning of UXR was when I tried to set up Mozilla DXR. It is a really nice and very useful tool, but the amount of dependencies makes it iimpossible to setup for a small project.
So, UXR takes the web frontend of DXR (with some modfications) and replaces the backend by a single, simple Python3 CGI script. The actual search backend can be a simple grep-like tool (at the moment the_silver_searcher, ag, or Russ Cox' great codesearch. The Clang-based indexer can still be used, but its output is simply stored and searched as CSV textfiles. No elasticsearch involved. No containers needed. No non-standard python dependencies.
Clone uxr, create a tree/ subfolder and populate it with code, and point a webserver at the CGI script.
python3 -m http.server --cgi
In uxr.py, you can change some settings and search backends. The default setup uses csearch (so install go and go get it), and needs an index of filenames. Create the indexes:
./make_cindex.sh
If you want the clang index, you have to create that first; refer to clang/make-index.sh.
- Only C/C++ really supported atm.
- The web frontend can freeze even fast browsers for a while when rendering a large (5kLOC) file.
- Syntax highlighting together with match highlighting is compilcated and broken.
- The precise search rules involving the clang index are not fully clear (normally all terms are
andconnected, but the index matches are in generalored into the results) codesearch/cindexseems to fail on huge indexing CSV files. Therfore,agis currently used for the index which does not scale well.- The UI is sometimes misleading, due to being ripped off its backend.
This is based on DXR, released by David Humphrey under the MIT license. Use the code added here under the same license.