At 12:24 PM 1/4/01 -0500, Charles Rich wrote:
>Paul,
>
> The revision you made to jtags below, in which the tagging process
>is broken into three passes, has an undesirable side-effect when the
>resulting TAGS file is used for tags-search or tags-query-replace.
>
> The problem is that each file in the source directory appears _three_
>times in the TAGS file, which means that each file will be visited three times
>by tags-search, etc. (I tested this). This is very inconvenient if you want
>for example, to review all uses of a particular identifier or replace
>only some occurrences of a pattern.
>
> I am not very familiar with xargs, but I wonder if there is some other way
>around the buffer problem you cite below.
>
It has been a long time since I made this change, but I think it was a
response to a problem that arose when I first tried to tag the JDK source
files several years ago. The JDK source hierarchy is huge and attempting to
tag it recursively in a single pass, using find and xargs, failed, at least
on Windows, because of some limitation on Cygwin xargs buffer capacity. I
was then and still am not very versed in either xargs or tags usage.
However, I've noticed that here at the Mathworks we use a hierarchical
tagging scheme for tagging large source hierarchies, i.e., a hierarchy of
tag files that mirrors the source hierarchy rather than one huge tags file
that contains tags for the entire hierarchy. So it seems to me that if one
used this approach it should be possible to tag each directory in a single
pass instead of multiple passes.
So some possible solutions to the problem are:
1) If it is possible to set xargs buffer capacity, up the capacity and
restore jtags to a single pass approach.
2) Disregarding the issue of xarg capacity, write a single-pass version of
jtags for use by people who have small hierarchies or who use a tag file
hierarchy to tag large source hierearchies where each tag file is small.
3) Create a Java program (or beanshell program) that does the same thing as
jtags, i.e.. recurse through a source hierarchy to build a single tags
file. This would avoid the find/xarg memory capacity issue.
- Paul
> Regards, and thanks for a _great_ JDE!
>
>-CR
>
># Revision 1.5 1997/12/03 03:31:29 kinnucan
># Divided tagging process into three passes
># through the source hierarchy to avoid overflowing
># the xargs buffer.
>
>--
> Charles Rich | Mitsubishi Electric Research Laboratories
> 617-621-7507 phone | 201 Broadway
> 617-621-7550 fax | Cambridge, MA 02139
> [EMAIL PROTECTED] | http://www.merl.com
