0. Clone source code:¶
$ git clone --depth=1 https://github.com/approach0/search-engine.git
The following instructions assume you have cloned this project in to directory
1. Install system dependencies¶
Other than commonly build-in libraries (pthread, libz, libm, libstdc++), ther are some system dependencies you may need to download and install on your system environment:
For Ubuntu 16.04 LTS (xenial), issue the following commands to install above dependencies:
$ sudo apt-get update $ sudo apt-get install g++ cmake $ sudo apt-get install bison flex libz-dev libevent-dev libopenmpi-dev
On Debian 9 for example, you may also need to install
$ sudo apt install libxml2-dev
2. Download and build Indri¶
This project currently relies on Lemur/Indri library to provide full-text index functionality (i.e. index writer and reader). To combine math and full-text search, we merge the results from Indri index reader and those from our math search, weight them using some combined score schema.
Lemur/Indri is not likely to be in your distribution’s official software repository, so you may need to build and manually specify its library path.
$ cd ~ $ git clone https://github.com/approach0/fork-indri.git ./indri $ (cd indri && ./configure && make)
If Indri reports
undefined reference to ... when building/linking, install that library and run configure again:
After installing the zlib-devel package you must rerun configure so that it correctly finds it and adds the library to the ld command.
Also, if you build Indri with newer version of gcc, it may produce segmentation fault when doing text indexing. Please refer to Appendix to resolve the issue.
3. Download CppJieba¶
Our indexer/searcher is able to handle English and Chinese document. Here CppJieba dependency enables Chinese term segmentation. Although we have option to choose whether or not to invoke this functionality when calling indexer/searcher programs, CppJieba is still required at compile time (at least for now) to successfully build this project.
However, there is no need to build CppJieba source code, only its C++ header files are required.
$ cd ~ $ wget 'https://github.com/yanyiwu/cppjieba/archive/v4.8.1.tar.gz' -O cppjieba.tar.gz $ mkdir -p ~/cppjieba $ tar -xzf cppjieba.tar.gz -C ~/cppjieba --strip-components=1
4. Configure dependency path¶
Our project uses
dep-*.mk files to configure most C/C++ dependency paths (such as CFLAGS and LDFLAGS). If you have installed above dependency libraries in your system environment, you can leave these
dep-*.mk files untouched. Otherwise if you compile and build dependencies locally, please modify
dep-*.mk files to point to your locally built library locations.
Before building, you need to invoke
./configure to locate all external dependencies:
$ cd $PROJECT $ ./configure --indri-path=~/indri --jieba-path=~/cppjieba
configure script also checks necessary libraries for building. If
configure outputs any library that can not be located by the linker, you may need to install the missing dependency before build.
make at project top level (i.e.
$PROJECT) will do the job.