Shell Programming and Scripting

I have some data that must have been two lines at some point. If the string in one cell is longer than 78 characters, the 79th character is a blank space. This was probably a return that was converted to a space somewhere along the line. This data is processed by some code that replaces blank spaces with underscrore. The result is an underscore at the 79th position that does not belong there.

For example:
N-{[3,4-dihydroxy-5-(hydroxymethyl)oxolan-2-yl]amino}(2-hydroxyphenyl)carboxam_ide

These files have tens of thousands to hunderds of thousands of rows, so this definately calls for a script. Is there a way I can check the 79th character of each string and if it is _, delete it? I can copy the column to a file if that would be easier, otherwise I will have to tell the script which column to process.

Here is a short sample of one of the columns that I need to process:

Code:

IUPACNAME
2-(2-cyclopropyl-2-oxoethylthio)-4,6-dimethylpyridine-3-carbonitrile
N-{[3,4-dihydroxy-5-(hydroxymethyl)oxolan-2-yl]amino}(2-hydroxyphenyl)carboxam_ide
butyl[(5-chloro-1-phenylpyrazol-4-yl)sulfonyl]amine
(1S,8R,9R)-10,14-dioxa-2,3,5,7-tetraazatetracyclo[6.4.1.1<9,12>.0<2,6>
(8S,8aR)-6-amino-2-benzyl-8-[2-(trifluoromethyl)phenyl]-1,2,3,7,8,8a-hexahydro_isoquinoline-5,7,7-tricarbonitrile
3-[N-(4-{[4-(3-carboxypropanoylamino)phenyl]sulfonyl}phenyl)carbamoyl]propanoi_c_acid
1-(phenylcarbonyl)-2-(3-pyridyl)-1,2,3,3a-tetrahydropyridino[2,3-e]indolizine-_3,3-dicarbonitrile
N-[(1E)-2-(3-phenylpyrazol-4-yl)-1-azavinyl]-2-[(3-fluorophenyl)carbonylamino]_acetamide
2-(5-acetyl-3-cyano-6-methyl(2-pyridylthio))-N-benzylacetamide

If the string is <79 characters, it can be skipped.

Thanks for the advice,

LMHmedchem

Code:

sed 's/\(.\{78\}\)_\(.*\)/\1\2/' x.dat
N-{[3,4-dihydroxy-5-(hydroxymethyl)oxolan-2-yl]amino}(2-hydroxyphenyl)carboxamide

Search for 78 instances of any character (and remember it), followed by an underscore, and remember the rest of the line.
If found, it will be replaced by the 2 remembered patterns, leaving the underscore out.

Thanks, that worked perfectly. Boy do I love sed, it's just amazing how quickly you can do complex text operations like this.

LMHmedchem

Becomes faster&simpler without the .* at the end of the regexp:

Code:

sed 's/\(.\{78\}\)_/\1/' x.dat