Search insde Word, PDF, XML and other files—installing and using crgrep
I am an editor in a business that uses Micro$oft products, but I want to be able to use the Linux CLI tools with which I am moderately familiar. In particular, I want to be able to grep Word documents, and that’s a problem because the new Word file format chops the text up and zips it up and hides it away. I googled and read a bit about crgrep (‘common resource grep’). Here is my experience so far.
Created a subdirectory c:\Users\username\installs\crgrep and downloaded the zip file into it. Worked in Cygwin, hence the forward slashes and dollar signs in the following. This could also be done through the GUI or in a PowerShell or CMD window. Choice is a wonderful thing.
$ unzip crgrep-1.0.5.zip $ cd crgrep-1.0.5/ $ vim INSTALL.txt
OK, so it needs java. Does it need the compiler (probably not, but check…). In the crgrep folder, typed:
$ grep -ir javac
Returned no results calling the javac compiler. So it looks like the program needs the runtime but not the development kit (JDK), so that’s good. It’s what you’d expect. Now, I have the wonderful ImageJ installed (works effortlessly in userspace), and it installs the Java runtime environment, JRE. Maybe I can use that.
Now, according to the INSTALL.txt file, the JAVA_HOME variable that crgrep wants points at something like
and my grepping told me that java.exe should be in %JAVA_HOME%\bin\java.exe
In Cygwin, my ImageJ tree looks like:
Which meant I needed to set JAVA_HOME to be C:\Users\username\installs\ij\ImageJ\jre (Windows-style path) (that is, the variable points the directory with the bin directory inside it, not the bin directory or the binary file itself.)
But first checked the version — needs 1.8.
$ cd ../../ij/ImageJ/jre/bin/ $ ./java.exe -version java version "1.8.0_112" Java(TM) SE Runtime Environment (build 1.8.0_112-b15) Java HotSpot(TM) 64-Bit Server VM (build 25.112-b15, mixed mode)
I installed in userspace (work computer, no root/admin access), so I went to my Windows account settings (given the various versions of Windows, I’ll assume a user can find their own account setting page) and (for Win 10; Win 7 is no doubt different) in the ‘Find a Setting’ box I typed ‘env’ for ‘environment variables’, and chose ‘edit the variables for your account’. Note that searching for ‘path’ turns up nothing. It’s a little trick!
Added an entry to the path:
(Path → Edit → New)
and created a new environment variable:
And exited everything, esp. the command line window, then opened it again, typed SET in a CMD (‘DOS’) window to see if the new variables were present, then tried the command:
H:>crgrep --help usage: crgrep [options]  crgrep: Common Resource Grep. -a,--text Process binary files or database columns as if they were text --color Alias for 'colour'. --colour Colour-highlight matched text ('always', 'auto' or 'never'). Default colour is red, see USAGE.txt for other colour settings. -d,--database Database grep (disables file search) -h,--help Help -i,--ignore-case Ignore case distinctions in matched text -l,--list List resources which produce a match by name. No content is searched. -m,--maven Include Maven POM file dependencies in search --mood Only include matching content expressing a specific sentiment; values include 'positive', 'negative' or 'neutral'. Ignored if -l specified. Requires model data; see INSTALL.txt --ocr Enable OCR text extraction from images; requires tesseract libraries. See INSTALL.txt -p,--password Password required to access a resource, optionally used with -u -P,--proxy Proxy settings for http access, specified as [:] -r,--recurse Recursive search into resources -u,--user User ID or username required to access a resource -U,--uri URI to specify a JDBC database resource -V,--version Print the version number of CRGREP to the standard output stream --warn Display all warnings to standard output -X,--extensions Enable one or more extensions; comma sep. list such as -Xdebug,trace If is not specified, or is '-', read from stdin Please report issues at https://bitbucket.org/cryanfuse/crgrep/issues
I want it for grepping Word files, so let’s see… yes, it finds ‘data’ in the test file, and outputs a nice clean stream:
H:>crgrep data text.docx text.docx:T:A key part of his research was the analysis of large datasets. As part of this he developed a software suite that included data modelling, reduction and correction techniques, and made of use the National Computing Infrastructure and other supercomputers. He enjoys the challenge of analysing and explaining complex data using words and carefully designed graphics. He likes Linux and the LATEXtypesetting system.
How about PDF? Converted the Word doc to PDF using the ‘Save as’ dialogue in Word. Then…
H:\>crgrep data text.pdf text.pdf:1:36:datasets. As part of this he developed a software suite text.pdf:1:37:that included data modelling, reduction and correction text.pdf:1:40:challenge of analysing and explaining complex data using
Different output because of how PDF and Word chop up the text, but instances found in both cases. No need to specify a file type or anything. I have not explored the command line options, but I am already finding the program useful — for example, when I want to find multiple instances of multiple expressions (say acronyms or references) in multifile projects.