could anyone tell me what these sed commands do and how I can do/insert them by hand.
sed -i '/^__func__.*__syncthreads/ {
i\
#if !defined(__CUDA__)
N
N
N
N
a\
#endif
}' $device_functions.h
sed -i '/^#define __device__ / {
i\
#if !defined(__CUDA__)
N
N
N
N
N
N
N
N
N
N
N
a\
#endif
}' $host_defines.h
This is what the programs do:
The first program searches for lines that match ^__func__.*__syncthreads and wraps that line and the next four following lines (the four N) in #if !defined(__CUDA__) and #endif by using i\ (immediate output) and a\ (appended output).
The second program searches for lines that match ^#define __device__ and wraps that line and the next eleven following lines in #if !defined(__CUDA__) and #endif.
This is done on the file specified by $host_defines.h and $device_functions.h, respectively, in an in-place manner (-i), so the files’ contents are changed.
Related
I would like to include tab as delimited new row to a file inp.txt.
This is the input produced by R:
inp <- 'AX-1 1 125
AX-2 2 456
AX-3 3 3445'
inp <- read.table(text=inp, header=F)
write.table(inp, "inp.txt", col.names=F, row.names=F, quote=F, sep="\t")
That´s what I am trying to do:
sed -i '1i The name\tThe pos\tThe pos2\' inp.txt
However, those three col names: 1- The name, 2- The pos, 3- The pos2 are not separated by tab in the output file. It just contain the \t string. Someone can help me here with the syntax?
Put the tab in a variable:
tab=$(echo "\t")
or
tab=$'\t'
Then you can use it in your sed script:
sed -i "1i The name${tab}The pos${tab}The pos2" inp.txt
I have a text file including lines in the form of:
(term1 x:a y:b (term2 z:c k:a))
I want to extract only terms from this line using command line utilities such as awk, grep, sed. i.e I want the result to be:
term1
term2
I have formed a regex matching the rest but the terms, but could not find a way to negate it.
(\()|( \()|( (.*?) \()|( (.*?)\)+)
How can I form a command extracting the every substring after '(' and before ' '?
Thanks
Try this:
sed "s/(\([^ (]*\)[^(]*/\1\n/g"
For example:
$ echo "(term1 x:a y:b (term2 (term3) z:c k:a) x (termX a:b ) )" | sed "s/(\([^ )]*\)[^(]*/\1\n/g"
term1
term2
term3
termX
I'm trying to replace two consecutive lines based on a pattern match, and would want this to repeat for the entire file. Here is the input file:
c aaaaa bbb
+ 0.1
c xxxx
c yyyy
+ 0.2
* c gggg
m eeeee hhhhh
+ 0.3
The command I tried is:
sed '/^c/{N;s/+/*+/}'
I expected to see a * prepended to each line beginning, but only those lines immediatlely following a c line:
c aaaaa bbb
*+ 0.1
c xxxx
c yyyy
*+ 0.2
* c gggg
m eeeee hhhhh
+ 0.3
what I actually get:
c aaaaa bbb
*+ 0.1
c xxxx
c yyyy
+ 0.2
* c gggg
m eeeee hhhhh
+ 0.3
Here, i see only the first occurrence of + (with previous line beginning with c) is getting replaced with *+. The second occurrence of + in the file is not getting replaced.
What am I doing wrong? How do I get the result I want: replacement happens in multiple consecutive lines in the file?
The problem you run into is that when a line that starts with c comes right after another line that comes with c, the N command in your code consumes it, and it isn't available for checking when you process the line that comes next.
Instead of reading ahead to see if the next line should be changed, I'd remember the last line and look back to see if the current line should be changed:
sed 'x; G; /^c/ s/+/*+/; s/.*\n//' file
This works as follows:
x # Swap pattern space and hold buffer. Because we do this here,
# the previous line will be in the hold buffer for every line
# (except the first, then it is empty)
G # append hold buffer to pattern space. Now the pattern space
# contains the previous line followed by the current line.
/^c/ s/+/*+/ # If the pattern space begins with a c (i.e., if the previous
# line began with a c), replace + with *+
s/.*\n// # Remove the first line (the previous one) from the pattern
# space
# Then drop off the end. The changed current line is printed.
sed -e 'H;$!d' -e 'x' -e ':cycle' -e 's/\(\nc[[:alnum:][:blank:][:punct:]]*\n\)+/\1*+/g;t cycle' -e 's/.//' YourFile
Posix version changing the whoe in max 2 internal cycle
load the file in memory (-e 'H;$!d' -e 'x')
Add the * in front of line starting with a + after a line starting with a c ( s/\(\nc[[:alnum:][:blank:][:punct:]]*\n\)+/\1*+/g)
do the same if occur in previous line ( :cycle and t cycle)
use a trick to insure starting with new line( H append current line to buffer also for first line so an extra new line as heading) (for first line with a c) and remove this at the end ('s/.//)
I have a report looks like this:
par_a
.xx
.yy
par_b
.zz
.tt
I wish to convert this format into csv format as below using sed 1 liner:
par_a,.xx
par_a,.yy
par_b,.zz
par_b,.tt
please help.
With awk:
awk '/^par_/{v=$0;next}/^ /{$0=v","$1;print}' File
Or to make it more generic:
awk '/^[^[:blank:]]/{v=$0;next} /^[[:blank:]]/{$0=v","$1;print}' File
When a line starts with par_, save the content to variable v. Now, when a line starts with space, change the line to content of v followed by , followed by the first field.
Output:
AMD$ awk '/^par_/{v=$0}/^ /{$0=v","$1;print}' File
par_a,.xx
par_a,.yy
par_b,.zz
par_b,.tt
With sed:
sed '/^par_/ { h; d; }; G; s/^[[:space:]]*//; s/\(.*\)\n\(.*\)/\2,\1/' filename
This works as follows:
/^par_/ { # if a new paragraph begins
h # remember it
d # but don't print anything yet
}
# otherwise:
G # fetch the remembered paragraph line to the pattern space
s/^[[:space:]]*// # remove leading whitespace
s/\(.*\)\n\(.*\)/\2,\1/ # rearrange to desired CSV format
Depending on your actual input data, you may want to replace the /^par_/ with, say, /^[^[:space:]]/. It just has to be a pattern that recognizes the beginning line of a paragraph.
Addendum: Shorter version that avoids regex repetition when using the space pattern to recognize paragraphs:
sed -r '/^\s+/! { h; d; }; s///; G; s/(.*)\n(.*)/\2,\1/' filename
Or, if you have to use BSD sed (as comes with Mac OS X):
sed '/^[[:space:]]\{1,\}/! { h; d; }; s///; G; s/\(.*\)\n\(.*\)/\2,\1/' filename
The latter should be portable to all seds, but as you can see, writing portable sed involves some pain.
I have files with entries of the form:
$$
y = x^2
$$
I'm looking for a way (specifically using sed) to convert them to:
\begin{equation}
y = x^2
\end{equation}
The solution should not rely on the form of the equation (which may also span mutiple lines) nor on the text preceding the opening $$ or following the closing $$.
Thanks for the help.
sed '
/^\$\$$/ {
x
s/begin/&/
t use_end_tag
s/^.*$/\\begin{equation}/
h
b
: use_end_tag
s/^.*$/\\end{equation}/
h
}
'
Explanation:
sed maintains two buffers: the pattern space (pspace) and the hold space (hspace). It operates in cycles, where during each cycle it reads a line and executes the script for that line. pspace is usually auto-printed at the end of each cycle (unless the -n option is used), and then deleted before the next cycle. hspace holds its contents between cycles.
The idea of the script is that whenever $$ is seen, hspace is first checked to see if it contains the word "begin". If it does, then substitute the end tag; otherwise substitute the begin tag. In either case, store the substituted tag in the hold space so it can be checked next time.
sed '
/^\$\$$/ { # if line contains only $$
x # exchange pspace and hspace
s/begin/&/ # see if "begin" was in hspace
t use_end_tag # if it was, goto use_end_tag
s/^.*$/\\begin{equation}/ # replace pspace with \begin{equation}
h # set hspace to contents of pspace
b # start next cycle after auto-printing
: use_end_tag
s/^.*$/\\end{equation}/ # replace pspace with \end{equation}
h # set hspace to contents of pspace
}
'
This might work for you (GNU sed):
sed -r '1{x;s/^/\\begin{equation}\n\\end{equation}/;x};/\$\$/{g;P;s/(.*)\n(.*)/\2\n\1/;h;d}' file
Prime the hold space with the required strings. On encountering the marker print the first line and then swap the strings in anticipation of the next marker.
I can not help you with sed, but this awk should do:
awk '/\$\$/ && !f {$0="\\begin{equation}";f=1} /\$\$/ && f {$0="\\end{equation}";f=0}1' file
\begin{equation}
y = x^2
\end{equation}
The f=0is not needed, if its not repeated.