From 3903f188544a1d1fe286fbe6acb9f465c4dd9a0c Mon Sep 17 00:00:00 2001 From: Albert S Date: Mon, 6 Jun 2022 14:49:24 +0200 Subject: [PATCH] Begin HACKING.md --- HACKING.md | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) create mode 100644 HACKING.md diff --git a/HACKING.md b/HACKING.md new file mode 100644 index 0000000..74d09a7 --- /dev/null +++ b/HACKING.md @@ -0,0 +1,31 @@ +# looqs - Hacking + +## Introduction +Without elaborating here, I hacked looqs because I was not satisfied with the state of desktop search on Linux. + +Originally a set of CLI python scripts, it is now written in C++ and offers a GUI made using Qt. While a "web app" would have been an option, I prefer a desktop application for something like looqs. I chose Qt because I am more familiar with it than with any other GUI framework. To my knowledge, potential alternatives like GTK do not include as many "batteries" as Qt anyway, so the job presumably would have been harder there. + +[CONTRIBUTING.md](CONTRIBUTING.md] contains the instructions on how to submit patches etc. + +## Security +The architecture ensures that the parsing of documents and the preview generation is sandboxed by [exile.h](https://github.com/quitesimpleorg/exile.h). looqs uses a multi-process architecture to achieve this. + +Qt code is considered trusted in this model. While one may criticize this, it was the only practical solution. looqs uses its serialization mechanism and other classes to communicate between the non-sandboxed GUI process and the sandboxed processes. + +## Database +The heart is sqlite, with the FTS5 extensions behind the full-text search. I definitely did not +want to run some heavy Java based solutions. I explored other options like Postgresql, I've discard them due to some limitations back then. + +Down the road, alternatives will be explored of course if sqlite should not suffice anymore. + +## File format support +The pdf library is libpoppler. Files such as .odt or .docx documents are opened with libquazip. The XML files in there are not parsed, +looqs simply strips the tags and that seems to work fine so far. Naturally, this is not the "proper way", so there is room for improvement maybe here. However, those file formats are not a huge priority for me personally. libuchardet does encoding detection and conversion. + +Naturally looqs won't be able to index and render previews for everything. Such approach would create a huge bloated binary. In the future, there will be some plugin system of some sorts, either we will load .so objects or use subprocesses. + +## Name +looqs looks for files. You as the user can also look inside them. The 'k' was replaced by a 'q'. Originally wanted my projects to have "qs" (for quitesimple) in their name. While abandoned now, this got us to looqs. + + +