The following steps will install Cheshire onto a linux machine from this repo. Currently only 64-bit versions of CentOS 6.5 and Ubuntu 14.04 are supported by these installation steps.

1. Grab dependency packages from the package manager: ``` centos> sudo yum install tcl tk tk-devel tcl-devel flex postgresql gcc kernel-headers libxml2 libxml2-devel postgresql-devel

or ubuntu> sudo apt-get install tcl tk tk-dev tcl-dev flex postgresql gcc linux-headers-$(uname -r) libxml2 libxml2-dev libpq-dev libc6-dev libxml2-utils 2. Download LLgen from http://cheshire.berkeley.edu/Downloads/LLgen_cheshire_v.tgz. Unzip using tar xzf LLgen_cheshire_v.tgz. Then edit the Makefile (if using CentOS 6.5), changing MANDIR=/usr/local/man to MANDIR=/usr/local/share/man then compile using make firstinstall. 3. Download the latest version of flex from Sourceforge. Untar the file, then modify flexdef.h replacing

```
#define MAXIMUM_NMS 31999
```

with

```
#define MAXIMUM_NMS 319999
```

then compile using the standard

```
./configure
make
sudo make install
```

Note, this requires sudo privileges for the current user.

1. Clone this repository or download and unzip the source tree.
2. Copy Makefile.centos65 or Makefile.ubuntu1404 to Makefile
3. Run make newbin to build Cheshire.

1. Download VIAF clusters XML file from viaf.
2. Unzip the file using gunzip filename.xml.gz
3. Build the associator file for cheshire:
   1. Make sure cheshire is in your path, using PATH=$PATH:~/cheshire/bin.
   2. Run buildassoc -q -s /path/to/viaf/filename.xml VIAFCluster.
      • The -q silences output
      • The -s skips other XML and non-XML code
      • VIAFCluster is the name of the XML element for which we are building an associator
   3. The associator will save a filename.assoc file for the XML, in the same directory as the XML file.
4. Create a database environment directory, mkdir DBENV.
5. Edit the config.viaf configuration file to match your setup. The following changes should need to be made:

<DBENV> /full/path/to/DBENV </DBENV>
<FILETAG> relative_path_for_filename.xml </FILETAG>
<FILENAME> /full/path/for/filename.xml </FILENAME>
<CONTINCLUDE> data4.cont </CONTINCLUDE>
<FILEDTD> viaf.dtd </FILEDTD>
<ASSOCFIL> filename.assoc </ASSOCFIL>

6. Index cheshire using index_cheshire -b config.viaf.
   • Check the logs (INDEX_LOGFILE) early using head or grep for the error NULL Document Tag Hash Table in comp_tag_list (in_components.c), which will cause no documents to be indexed.

Note: For the Aug 28, 2014 VIAF data dump, there were 26,793,357 records. However, the VIAF format changed from May 2014 to August 2014, therefore the viaf.dtd file must be updated to accommodate the newer format. I was unsuccessful at getting it correctly updated, so I used the older records, found on shannon at ~rlarson/DATA/VIAF/data4.

Cheshire indexing does not require much space, as it turns out. The VIAF records use 72GB (for the Aug 2014 set) or 86GB (for the May 2014 set). The associator file is approximately 200MB, with the indexes and BerkeleyDB database environment totalling approximately 30GB. The logfile generated from index_cheshire was 16MB in size. All of VIAF could be indexed, therefore, with around 120GB of disk space.

Cheshire indexing itself does not require much memory, however if using the -b option to index_cheshire, the sort command will require enough memory to sort the indexes. For example, indexing the VIAF records above requires up to 8GB of RAM.

1. Grab dependency packages from the package manager:

   centos> sudo yum install postgresql postgresql-contrib python-setuptools python-devel swig
   

or
ubuntu> sudo apt-get install postgresql postgresql-contrib python-setuptools python-dev swig 2. In the former repository, change directory to CheshirePy using cd CheshirePy. Then run the make commands: make clean && make. 3. Copy the compiled CheshirePy to its correct location, with the following commands:
sudo cp _CheshirePy.so /usr/lib/python2.7/lib-dynload/. sudo cp CheshirePy.py /usr/lib/python2.7/.