使用 sed 修复有问题的文本文件的行

Question

During a pipelined process in a custom fraimwork I am working on, there is a need to process files which are generated from a certain engine. But, the thing is that the file format for certain lines is kind of broken. Meaning that there is an inconsistent way of printing out the lines. Like this:

/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_A 
                                                             0           0        0.00 
/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_B
                                                             0           0        0.00 
/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_C 
                                                             1           1       100.00
/SLIGHTLY_SMALLER_NAME_OF_ITEM_D                             0           1        50.00
. 
. 
. 
.

Which I wish to transform to

/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_A        0           0    0.00                                                            
/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_B        0           0    0.00 
/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_C        1           1    100.00
/SLIGHTLY_SMALLER_NAME_OF_ITEM_D                             0           1    50.00

The problem here is that for a single entry the fields AB C are appearing after \n while for other entries (eg, D) the line is consistent.

By using a very helpful tool (eg, regex101 ) I've managed to build a Regular expression that covers and groups the lines. The regex is the following:

(\/.+)\n\s+([0-1]\s+[0-1]\s+.+\b)
-----  ---  ---------------------
  |     |            |
  |     |            |=> groups the secondary line containing the digits (the first two are only 0|1)
  |     |
  |     |=> new line along with all the whitespace untill the first digit 
  |
  |=> groups the first string-stream (ex: /VERY_LONG_NAME_...)

The thing is that I am trying to re-create the file by using (most probably in an erroneous way) sed as:

sed -r 's/(\/.+)\n\s+([0-1]\s+[0-1]\s+.+\b)/ \1 \2/' filename.txt

which of course it does not work as I expected. So am I doing something wrong here? Syntactically wise at least? Furthermore, I do not wish to modify the "CORRECT" lines, meaning the lines that are not "broken" into two lines. I just want to fix the 'problematic' ones

Answer 1

With awk and column :

awk 'NF==1{x=$0; getline; $0=x OFS $0} {print}' filename.txt | column -t

If current row has only one column ( NF==1 ) then save complete row to variable x and read next row ( getline ) and concat last row ( x ) with output field separator ( OFS ) and current row ( $0 ) to new current row ( $0=x OFS $0 ).

Output:

/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_A  0  0  0.00
/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_B  0  0  0.00
/VERY_LONG_NAME_FOR_A_SPECIFIC_ITEM_IN_THIS_PROCESS_C  1  1  100.00
/SLIGHTLY_SMALLER_NAME_OF_ITEM_D                       0  1  50.00

See: 8 Powerful Awk Built-in Variables – FS, OFS , RS, ORS, NR, NF , FILENAME, FNR