Documentation for the DSE DELPHI Search Engine

This is a short introduction explaining some features of the DSE DELPHI Search Engine and its management.

  • Structure of the engine;
  • Use of the engine;
  • Management of the figure archive;
  • Insertion of Paper figures;
  • Structure:

    The DELPHI Search Engine is based on a nph-cgi script written in Perl by the author (Marco Ganz) and running under the web-server at This script parses the queries, retrieves the figures in the archive and outputs a virtual page with the matching results. The figure archive is on afs under the directory /afs/ Here are placed two files used by the script ("indexes" and "keylist") and a set of directories corresponding to the different groups of images, containing the figures, their captions and gifs. Moreover for each of this group there is a file "group.index" containing, for each figure, the path to retrieve it (relative to "/afs/") and the related keywords.


    The DELPHI Search Engine may be invoked from the "Keywords List Page" or directly from a previously generated "results" page. In both cases a query can be specified in the "Query" text field in a boolean form. All the keywords took from the displaied list, parenthesis and the boolean operators "AND", "OR" and "~" are accepted. The query can be submitted using the "submit" botton or pressing <CR> after insertion. Before submission it is possible to choose the level of importance of the images that are to be retrieved: if "high" is selected from the menu all the "low importance" images (images with a "VAL0" keyword associated) are skipped, otherwise ("low" is specified) all the images are scanned for matching. The output page contains the unordered list of the "small gifs" of the retrieved figures and their captions. It is possible to have a "Ghost View" of the postscript of the figure clicking over the relative miniature. At the end of the page is a link to go back to the"Keywords List Page".

    Management of the figure archive

    The figure archive is contained in the "/afs/" tree of the afs file system. It can be managed by the "delfigs" user or any member of group xx:www or xx from any afs client machine. To insert new figures in the archive it is sufficient to copy its postscript file, its description file and its miniature (these must have the same name with the following extensions: .[e]ps, .txt (or .html) and .small.gif respectively) into the directory of the relative group. Then a new reference must be inserted into the group.index file: the new line must start with the path of the figure relative to /afs/ followed by the list of the keywords associated to it separated by ; (semicolon). This list must start and end with a semicolon. If the figure has a low level of importance the VAL0 keyword may be specified too, in order to let the engine skip the image when the "high" level search is specified. When the image is correlated to the preceding one in the index, after the ";" terminating the keywords list may be specified a A character instructing the engine to display this image whenever retrieved without the separation line from the preceding image. When the second line of the image description (image.txt) is a newline ("\n") the first line is interpreted as a title and so displaied in bold characters. When a group archive is created for the first time the reference to its group.index file has to be added to the "/afs/" file to enable the engine to retrieve it (always with path relative to /afs/

    Example: The group ABC has a figure directory called /afs/ To insert the Fig image in the archive the files Fig.eps, Fig.txt and Fig.small.gif, are to be copied into the group directory and the following line is to be added in the ABC.index file present in the groupdir:
    If the figure is with low importance the VAL0 keyword can be specified. If the image is to be displaied without interruption from the one retrieved before it the A character is to be specified at the end of the line so it becomes: ABC/Fig;KEY1;KEY2;...;KEYn;A

    Insertion of paper figures

    To insert the material relative to a paper stored into the "vxdel1" papers archive an ad hoc script has been created so that all this operation can be performed automatically. The script is contained into the "tools" directory in the home directory of the "delfigs" user and is called ep. The script is invoked together with the number of the paper that has to be archived (ATT.N: if the file document.tex of the paper is in vxdel1 in the directory paper0999 the number to pass to ep is 0999 and NOT 999!). If everything goes well a new directory is created into /afs/ with all the usual files and the reference to the new index is added to the "indexes" file.
    The idea on which "ep" is based on is that of having "ONE POSTSCRIPT ONE FIGURE" so when more than one figure is in the same "begin{figure}...end{figure}" block only the first one is retrieved;
    "ep" automatically associates to the extracted figure the "PAP0999" keyword; further keywords for that figure can be specified by the author directly in the "Document.tex" file: following the caption and within the begin{figure}...end{figure} block, with the following syntax: %KEYS{ key1; key2; ...; keyN} ;

    Example: To archive the paper 0999 it is sufficient to give a ~delfigs/tools/ep 0999 command. All the files relative to the paper should be automatically created and the index too.


    Comments and questions send to

    This document was written by Marco Ganz in 1996,
    updated by Jiri Chudoba.

    Last modification: