The concepts of planning I offered do not derive from the workflow, indeed may help DETERMINE the workflow. They depend upon analysis of the DATA flow as derived from the statement of the problem, data input, desired transformations, and data output, and not at all on the tools used. There is no circular dependency. I see why you think there may be, but you if you consider it as a programming problem there is none.

BASH can do this, PERL is almost DESIGNED to do this (and most things, it is one of the most versatile tools), or you can do this in python, Pascal, c, assembler, BASIC, FORTRAN, COBOL ... : the list is almost endless.

The OP needs to consider what he WANTS to use (BASH is an expressed preference, see the title of the thread) so we can assume that over awk, perl, sed, etc.

Bash can, using only internal commands and shell features, read files, compare and use strings, create files and folders, all of the critical pieces are there for this problem without calling awk, sed, grep, perl, or other external tools. He may, on consieration, wish to use one or more external tools as a 'helper' for his script, but it can be done without those.
.

i think by circular dependence he meant that it would be hard to draw a diagram without knowing the feature set of awk, for example ?

That makes sense. But it starts by assuming that you will first look at your tools to decide what you want to do. That seems backwards. First define properly what you want to do THEN look at your tools and find the right ones.

Both plans are partly out of line here. The question was how to do it in BASH, so awk, sed, etc. are off the table. Properly defining what you want to do is still valid.

I am looking forward to seeing what the first cut looks like. There are so many good ways to approach this.

Yeah. That's about it. Also since this is a newbie subforum, the word "bash" often ends up meaning "anything except GUI" Hence the pointer to awk and Perl

I needed it using awk, did not know it wasn't a part of bash. Below code worked as per need.
Thanks all.

awk -F' ' '{fn=$1".log";print>fn;close(fn)}' \FS='\n' OFS='\n' RS='---+\n' ORS='' filename

Great. You could even trim the file names to just the date using sub()

See "man awk" for the full manual.

awk is acessible via bash but not part of it. bash is just one of the interfaces to the system. You have bash, zsh, ksh, and dash on the one hand for text. On the other hand for graphics, you have Unity, XFCE, KDE, LXDE, and Openbox, just to name a few from either category.

Thanks. But, i am looking for the file format to be <filename_YYYYMMDD.log>. Can anybody help on this please?

Yes, check "man awk" for info on sub(), gsub(), match(), or substr() to see how far you can get with rearranging the existing numbers. Probably the latter two are most relevant. Maybe do not append the '.log' part to fn until you've got the first part of the file name settled.

As things get more complex, the problem becomes more appropriate for perl. But it should be doable within awk still.

Bash has length and string parts manipulation built in.

$ MYDATE="01-01-1970"; echo "MYDATE contains "$MYDATE" and has "${#MYDATE}" characters."; echo "YYYY="${MYDATE:6:4}" MM="${MYDATE:3:2}" DD="${MYDATE:0:2};

Note to self, buy less tea. In that example the offsets assume a DD-MM-YYYY format. Apparently I woke up british or something this morning.

if you know the pattern is going to be a filename[underscore]Numbers indicating a date. then why not search for the underscore then chop it off giving you the date to do with what you will?

or are you saying something else

http://www.theunixschool.com/2012/06...file-into.html

OK, that is not the file format, the is the file NAME format. That is easy enough to put together in BASH, it should be doable easily in AWK (or a combination) but I do not AWK enough to say how.

if you can use bash script then I got a solution for ya. but if it got a be awk then I'm out too, I do not use that.