Chapter 18. RIS and risx format specifications

Table of Contents

The RefDB RIS implementation
The risx DTD

In general, RefDB tries to stick as closely as possible to the RIS format specification of the Reference Manager software version 8 (this is the most recent version I was working with). This chapter briefly recalls the general syntax of RIS and risx files and the few differences in the implementation between Reference Manager and RefDB.

The RefDB RIS implementation

Import from Reference Manager and similar bibliographic databases as well as export should work in general, although it is not always loss-free due to the varying degree of support for the full RIS specification. This chapter briefly summarizes the differences between the Reference Manager RIS specification and what RefDB uses.

  • In contrast to the mainly Windows-based bibliographic databases, RefDB expects the RIS files with UNIX line endings (LF = 0x0A). Most online sources allow to download the files in this format. The RefDB package also contains a refdb_dos2unix shell script which converts the line endings from DOS format to UNIX format (you may use your favourite perl or whatever program instead, of course).

  • In Reference Manager, the number of authors/editors/series authors per reference is limited to 255. Although this may not have any practical consequences, RefDB does not have this restriction.

  • The same holds true for keywords. In RefDB the number of keywords per reference is not limited.

  • Some of the tags defined in the RIS format specification have synonyms. RefDB does not use these randomly upon output of RIS datasets, but uses a defined subset. This subset may be a different one than Reference Manager uses. This should have no practical consequences for the data integrity, although this means that after shuffling datasets between the two databases different synonymous tags may be used.

  • The BT/T2 field is limited to 16KB in Reference Manager. It is limited to 64KB in RefDB.

  • A couple of fields in a Reference Manager database have either “no practical length limit” or the length limit is not specified at all. As these field lengths are not exactly given in SI units, I can just list what the corresponding limits in RefDB are: PY/Y1, Y2 are limited to 255 characters starting from the first slash (i.e. excluding the year information), TI/T1/CT, IS/CP, BT/T2, T3, AD are limited to 64KB, AB and N1 are limited to 16MB. I assume that none of these poses a “practical length limit”. If you should plan to exceed these limits, you can of course recompile refdbd and use up to the maximum field length that the database server offers (e.g. 4GB in MySQL).