Haha, wow... I'm sorry I seemed to have sparked quite the discussion :-p I certainly appreciate all the help!
I am indeed an uber noob, and I have actually zero experience with awk. Although it seems to make sense to separate the data into "|" delimited fields, the sed command seems much simpler and easier for me to read (even though it is still a bit cryptic.... regex is seriously the bane of my existence). Plus, the case previously mentioned in which I might have more than one tcolor field, for example, would not be an issue.
That being said... unfortunately, doing the following:
Code:
grep "title=Boston" datafile | sed 's/tcolor=[^|]*/tcolor=blue/'
...does not seem to change the info in the datafile itself.
See, I have an html form that allows the user to enter in these fields into text boxes.. then they are added to the datafile via a .cgi script. If the user goes back and enters in new info for a data item who's title already exists in the datafile, then the new values are meant to replace the old.
So, if the datafile is as follows:
Code:
title=Boston|tcolor=green|desc=Large city in New England|url=http://www.boston.com
title=New York|tcolor=gold|desc=The largest city on the East Coast|url=http://www.newyork.com
title=Dallas|tcolor=silver|desc=A sprawling, cheery Texas city|url=http://www.dallas.com
and the .cgi script passes in the following parameters in the query string:
Code:
title=Boston&titlecolor=blue&descrip=&url=
Once the query string is parsed, what would be the best way to go about testing which parameter is not an empty string, and only updating that field (in this case the tcolor field in the $title line with $titlecolor) in the datafile?
...I hope that all makes sense, haha. Again, thank you for the help!
i assume you are going to parse the above string, so once again, the distinct field delimiter you can spot easily, its "&". split on "&" , then go through each item and check is there is anything after the "=" sign, if there is nothing, then its a null string. OR you can split on "=" , and make sure that the next field doesn't start with "&".
Location: Asia Pacific, Cyberspace, in the Dark Dystopia
Posts: 19,118
Thanks Given: 2,351
Thanked 3,359 Times in 1,878 Posts
When a poster, who is not posting homework of course, asks for a reply in sed, it is generally best not to try to twist the arm of the original poster to a solution using awk or some other tool.
The poster asked for an sed solution and has commented they are not familiar with awk. So, let them have a solution that works for them.
I tend to agree that this has turned into a hypothetical discussion and does not address the needs or requirements of the original posters question.
Unless there is some reason the original poster wants to switch to awk, let's give him a solution in sed, which is what they asked for in the first place.
Hi, you could combine a function and a sed statement to create a way of changing a field value based on a field name. E.g.
Code:
replace_field()
{
sed "/^$1/s/\(${2%=*}=\)[^|]*/\1${2#*=}/" infile;
}
replace_field "title=Boston" "desc=bla bla"
Code:
output:
title=Boston|tcolor=green|desc=bla bla|url=http://www.boston.com
title=New York|tcolor=gold|desc=The largest city on the East Coast|url=http://www.newyork.com
title=Dallas|tcolor=silver|desc=A sprawling, cheery Texas city|url=http://www.dallas.com
This probably looks quite cryptic but it is based on the previous examples with the values replaced by $2 (and parameter expansions). The function takes the first key to locate the line on which certain fields need to be replaced.
We can expand this to take more fields and replace the $2 expansions with field names so it becomes a bit more readable..
Code:
replace_fields()
{
key="$1"; shift
sedstr="/^$key/{"
while [ $# -gt 0 ]; do
field="${1%=*}"
val="${1#*=}"
sedstr="${sedstr}s/\b\($field=\)[^|]*/\1$val/;"
shift
done
sedstr="$sedstr}"
sed "$sedstr" infile
}
replace_fields "title=New York" "tcolor=green" "desc=blabla"
Output:
Code:
title=Boston|tcolor=green|desc=Large city in New England|url=http://www.boston.com
title=New York|tcolor=green|desc=blabla|url=http://www.newyork.com
title=Dallas|tcolor=silver|desc=A sprawling, cheery Texas city|url=http://www.dallas.com
If you want to edit the file directly
you could use:
Code:
sed -i "$sedstr" infile
if your sed supports it or
Code:
sed "$sedstr" infile > /var/tmp/infile.tmp.$$ && mv /var/tmp/infile.tmp.$$ infile
It builds a sed statement based on the inout fields that are being provided. Unfortunately I haven't been able to make it simpler than this.
Also, it has to be said that this will probably only work in very low intensity situations. If more users will be creating updates simultaneously then problems may occur and you will have to create some sort of locking-and-wait mechanism. If it is getting more serious you will probably be better off using a simple database.
Last edited by Scrutinizer; 12-12-2009 at 04:49 PM..
Hi everyone,
I need very simple sed command to change a parameter in a text file.
I have a line in this text which is like
set xx 0.5
A program reads this file and does some algebraic calculations. So to make a parameter scan I need to change the value of xx. I thought I can do... (7 Replies)
How can you use sed with a line of code that reads:
67899:Bill:Williams:Maple Dr.:45908600
Let us say we want to replace Maple Dr. with Oak St. (1 Reply)
Hello,
I have two files. File1 is normal txt file and File2 contains list of line numbers.
e.g. File2:
3
6
9
.....
I need to replace a character in File1 in lines (taken from File2). For that I am using a "for" loop:
for i in $(cat File2)
do
sed "$i s/Y/N/" File1
done
but my... (3 Replies)
I have a file with some numbers having single quotes around them which I want to remove.
i.e. '923930' -> 23930
If it can be done without using sed thats fine. I have tried with sed but can't think how to replace this pattern on only the numbers (13 Replies)
Hi I am trying to do a text insertion in a text file at a particular line number in a shell script. However its not working.
sed '122i\
> for j in \`echo $MyList\` ; do perl -pi -e\'s#01\/01\/2009#01\/01\/2011#\' $j ; done' $HOME/MyScript.ksh
The Actual line to be inserted at line 122... (5 Replies)
Hi guys,
Can u please help me to replace (-) with (/) in a file containing no of records using "sed " command in unix.
thanks in advance.
subhendu (5 Replies)
Hi ,
I am stuck up in the below scenario:-
I need to read a file name (eg A.txt) name frm another file (eg B.txt) and then I need to search for a particular expression in A.txt and substitute it with another expression.
How can I use SED inside SHELL Scripting and command prompt as well to... (1 Reply)
Hi ,
I am stuck up in the below scenario:-
I need to read a file name (eg A.txt) name frm another file (eg B.txt) and then I need to search for a particular expression in A.txt and substitute it with another expression.
How can I use SED inside SHELL Scripting and command prompt as... (2 Replies)
Hi,
I have a set of files containing strings like I.TEST1_TEST2 or B.ESSA_ESSB for example.
Does somebody know how to substitute these strings whith the same name and an extension "_V1" (ie. I.TEST1_TEST2_V1) using sed command or else ? (3 Replies)
I know we can substitute a string using sed but how?
For example: sed 's/(old variable)/(new variable)/ details.dat
Am I suppose to put $old variable or whatever? Because I tried many times, it didnt work by putting $old variable. Am I suppose to enclose it with "" or ''?
Please help (3 Replies)
