ANT vs Make

[Leslie Chaim]

Is there any article which does a direct comparison between ANT vs Make. I've read the article from "The Saint". I've also read an article in the Java developer Journal and I am ready to dive in.
As a Make user, I already understand the concepts of targets and dependencies. I think I just need to adjust to ANT's syntax. What do you think?
Thanks,
Leslie

[Kirk Pepperdine]

Hi Leslie,
I don't know of any direct comparison between ANT and make. Pick a problem that you need to solve and dig in. If you're used to make and understand XML, you should do just fine.
Kirk

[Carol Enderlin]

I found one comparison:
Comparing ant and make as java build tools
Maybe it's old, it lists one of the cons for ant is that you compile all classes every time. That doesn't have to be true.
[ March 10, 2003: Message edited by: Carol Enderlin ]

[Andy Bowes]

Hi Carol
I think that exactly the opposite is the case. Ant is v smart when it comes to identifying which classes need to be compiled and which ones in a jar file need to be repackaged etc.
This results in a significant improvement in the performance of the build as opposed to relatively dumb MAKE files.
IMHO, Ant totally simplifies the development process even though it can take a bit of time and thought to break down the operations into the appropriate tasks and identify the dependencies between these tasks. One of the key features of Ant is it's open API which allows you to create your own custom Tasks.
Our current implementation includes an Ant script that is automatically run every night which performs the following operations:
1. Check out the latest source code from our version control package.
2. Compile the source code
3. Build the deployment packages.
4. Start our test server
5. Deploy the packages to the server.
6. Run a whole suite of test scripts and send emails to appropriate people detailing any failures.
7. If the whole application builds and tests successfully then deploy the application to our integration server.
Don't fancy writing a MAKE file to do that

[Leslie Chaim]

Originally posted by Andy Bowes:

...
Don't fancy writing a MAKE file to do that

Well, some have done that .

[Leslie Chaim]

Well, here is something which I have done using Make. It's very complex and a bit contrived but I was wondering how would this be done in ANT.
First, a brief overview of my problem:
I wanted to build prog (which is written in C) based on all of its dependencies. However, some commands (or tasks) should always run and Make should decide whether to rebuild prog. I made use of The double colon of Make.
The task involved of building two dynamic header files based on information in an Oracle database! This allowed "prog" to be built based on meta-data. One rule always ran which refreshed (or replaced) the header file if it was necessary. If there was a change in the header file then Make would run the second rule.
I really had to do some arm twisting (as you will see) to get this done. I am just wondering how much easier this would have been with ANT.
Here's the snippet from the Makefile:

CFLAGS = +w1 -Aa -g -I/usr/local/include -D_PROTOTYPES
 
C_LIBS = -lm
 
include /usr/local/include/Oracle.libs #BTW, what's ANT equivalents?
 
HEADERS =       \
 prog.h         \
 dyn_defs.h     \
 map_lst.h
 
OBJECTS = all the .o files and what ever?
 
prog:: dyn_defs
 
prog:: prog.c $(HEADERS) $(OBJECTS) $(OTHER_LIBS)
        $(CC) prog.c -o prog $(CFLAGS) $(OBJECTS) \
        $(OTHER_LIBS) $(ORACLE_LIBS) $(C_LIBS)
dyn_defs:
        @update_header create_define_file dyn_defs
        @update_header create_map_file   func_lst
 

<code>update_header</code> is a Shell script:

 
script=$1
target_file=$2
 
if var=$($script.sh | diff - $target_file.h)
then
   echo "There were no changes in $target_file.h."
else
   echo There were differences in $target_file.h, Here ...
   echo "\n\n===>$var<==="
   echo updating $target_file.h
 
   #co -q -l $target_file.h
 
   $script.sh > $target_file.h
 
   #ci -u -q -m"$var" $target_file.h
fi

As you can see update_header regards to the first param as a script. Both <code>create_define_file.sh</code> and <code>create_map_file.sh</code> do the same thing: they query the database and return a complete header file.
I will just show the more complicated of the two: create_map_file.sh

# sql2file is a homemade tool to query the database with ease.
# It returns the data from a select statement in tilde("~")delimited fields.
# Here we pipe it to Perl and create our map_file.h
sql2file -t -u scott/tiger "
select func_ptr, func_name
  from meta_data_list
 order by func_name "  -  /dev/null |
 perl -w -pe '
BEGIN {
print <<MAP_LST_HEADER;
#ifndef MAP_LST_INCLUDED
#define MAP_LST_INCLUDED
 
/**
*
*  This structure contains the function names and pointers to them.
*
*  The last element of the list contains a null pointer and a function
*  name that guarantees that this element will collate to the bottom of
*  the list.  Do not modify or move this entry!
*
*  As of Wed Nov  8 16:04:59 EST 2000 this file will now be automated
*  through a process by extracting the data from an Oracle table.
*  Furthermore, this will be moved to the Makefile. And the make(1) command
*  will ensure that the entries are upto date!
*  Therefore, DO NOT MANUALLY EDIT THIS FILE. Instead all entries or modification
*  should be made in the "meta_data_list" table.
*
**/
 
static MAP_FUNCTION_LIST_STRUCT func_list [] =
{
 
MAP_LST_HEADER
} # End of Perl BEGIN block
 
    chomp;
    # Split the two fields and print them nicely
    s/([^~]*)\~(.*)/sprintf "      %-30s \"$2\",\n", "$1,"/e;
 
END { print <<MAP_LST_TRAILER;
      NULL,                          "ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ"
};
 
#endif  /* MAP_LST_INCLUDED */
MAP_LST_TRAILER
} # End of Perl END block
' # End of shell command (Yes this is crazy but fun)

A lot of arm twisting I think so! How would this be achieved with ANT?
Thanks,
Leslie
[ March 12, 2003: Message edited by: Leslie Chaim ]