Search and capture multiple groups between two patterns [duplicate] - bash

Question

This question already has answers here:
How to print lines between two patterns, inclusive or exclusive (in sed, AWK or Perl)?
(9 answers)
Closed 5 years ago.
I need to capture the content of my alert log for lines starting with ORA-01555 and their related SQL statements.
The file content is below:
Sat Oct 21 12:11:40 2017
Thread 1 advanced to log sequence 143 (LGWR switch)
Current log# 5 seq# 143 mem# 0:
Sat Oct 21 12:12:22 2017
Thread 1 cannot allocate new log, sequence 144
Private strand flush not complete
Current log# 5 seq# 143 mem# 0:
Sat Oct 21 12:12:22 2017
ORA-01555 caused by SQL statement below (SQL ID: 1sfgms7z66zrs, SCN: 0x0401.1e45bbc4):
Sat Oct 21 12:11:48 2017
CREATE INDEX AAD_SAA_ADB_RESULTS_AUD_GTI1 ON AAD_PS_SAA_ADB_RESULTS_AUD_GT (
rpt_date
)
Sat Oct 21 12:20:56 2017
Thread 1 advanced to log sequence 153 (LGWR switch)
Current log# 6 seq# 153 mem# 0:
+REDO_VOL1/DSS1STG/ONLINELOG/group_6.262.957710915
Current log# 6 seq# 153 mem# 1:
+REDO_VOL1/DSS1STG/ONLINELOG/group_6.263.957710917
Sat Oct 21 12:21:11 2017
ORA-01555 caused by SQL statement below (SQL ID: 9ad5awvfvsfd9, Query
Duration=6444 sec, SCN: 0x0401.1e53de65):
Sat Oct 21 12:21:11 2017
SELECT a.EMPLID, a.TEST_ID, a.TEST_COMPONENT, a.SCORE
FROM
ODS_PSE.PS_STDNT_TEST_COMP a,
DSS_RDS.IR_STU_TST_SCR_SAT_VLD_DT_ST b
where
a.TEST_COMPONENT in ('VE','MA')
AND a.EMPLID = b.PRSN_UNIV_ID
AND ROUND(a.SCORE) = a.SCORE
AND a.TEST_ID = b.STU_TST_CD
AND a.TEST_DT = b.TEST_DT
AND a.LS_DATA_SOURCE = b.LS_DATA_SOURCE
UNION
SELECT a.EMPLID, a.TEST_ID, a.TEST_COMPONENT, a.SCORE
FROM
ODS_PSE.PS_STDNT_TEST_COMP a,
DSS_RDS.IR_STU_TST_SCR_SAT_VLD_DT_ST b
where
a.TEST_COMPONENT in ('ERWS','MSS','MT','RT','WLT')
AND a.EMPLID = b.PRSN_UNIV_ID
AND (
(ROUND(a.SCORE) = a.SCORE and a.TEST_COMPONENT in
('ERWS','MSS','RT','WLT'))
or (a.SCORE BETWEEN 10 AND 40 and a.TEST_COMPONENT in ('MT'))
)
AND a.TEST_ID = b.STU_TST_CD
AND a.TEST_DT = b.TEST_DT
AND a.LS_DATA_SOURCE = b.LS_DATA_SOURCE
Sat Oct 21 13:05:01 2017
Thread 1 advanced to log sequence 173 (LGWR switch)
Current log# 5 seq# 173 mem# 0: +REDO_VOL1/DSS1STG/ONLINELOG/group_5.257.957710909
Current log# 5 seq# 173 mem# 1: +REDO_VOL1/DSS1STG/ONLINELOG/group_5.256.957710911
The content I want to capture is:
ORA-01555 caused by SQL statement below (SQL ID: 1sfgms7z66zrs, SCN: 0x0401.1e45bbc4):
Sat Oct 21 12:11:48 2017
CREATE INDEX AAD_SAA_ADB_RESULTS_AUD_GTI1 ON AAD_PS_SAA_ADB_RESULTS_AUD_GT (
rpt_date
)
ORA-01555 caused by SQL statement below (SQL ID: 9ad5awvfvsfd9, Query
Duration=6444 sec, SCN: 0x0401.1e53de65):
Sat Oct 21 12:21:11 2017
SELECT a.EMPLID, a.TEST_ID, a.TEST_COMPONENT, a.SCORE
FROM
ODS_PSE.PS_STDNT_TEST_COMP a,
DSS_RDS.IR_STU_TST_SCR_SAT_VLD_DT_ST b
where
a.TEST_COMPONENT in ('VE','MA')
AND a.EMPLID = b.PRSN_UNIV_ID
AND ROUND(a.SCORE) = a.SCORE
AND a.TEST_ID = b.STU_TST_CD
AND a.TEST_DT = b.TEST_DT
AND a.LS_DATA_SOURCE = b.LS_DATA_SOURCE
UNION
SELECT a.EMPLID, a.TEST_ID, a.TEST_COMPONENT, a.SCORE
FROM
ODS_PSE.PS_STDNT_TEST_COMP a,
DSS_RDS.IR_STU_TST_SCR_SAT_VLD_DT_ST b
where
a.TEST_COMPONENT in ('ERWS','MSS','MT','RT','WLT')
AND a.EMPLID = b.PRSN_UNIV_ID
AND (
(ROUND(a.SCORE) = a.SCORE and a.TEST_COMPONENT in
('ERWS','MSS','RT','WLT'))
or (a.SCORE BETWEEN 10 AND 40 and a.TEST_COMPONENT in ('MT'))
)
AND a.TEST_ID = b.STU_TST_CD
AND a.TEST_DT = b.TEST_DT
AND a.LS_DATA_SOURCE = b.LS_DATA_SOURCE
Here is the grep I'm using but as you can see in the result it isn't capturing the content for each ORA-01555 line and its entire SQL statement. Not sure why my negative lookahead isn't quite working:
grep -Pzo 'ORA-01555.*\n(^.*[0-9]{2}:[0-9]{2}:[0-9]{2}.*)\n.*(?!.*[0-9]{2}:[0-9]{2}.*\n).+' alert.log
ORA-01555 caused by SQL statement below (SQL ID: 1sfgms7z66zrs, SCN: 0x0401.1e45bbc4):
Sat Oct 21 12:11:48 2017
CREATE INDEX AAD_SAA_ADB_RESULTS_AUD_GTI1 ON AAD_PS_SAA_ADB_RESULTS_AUD_GT (
ORA-01555 caused by SQL statement below (SQL ID: 9ad5awvfvsfd9, Query Duration=6444 sec, SCN: 0x0401.1e53de65):
Sat Oct 21 12:21:11 2017
SELECT a.EMPLID, a.TEST_ID, a.TEST_COMPONENT, a.SCORE

Answer 1

You can try this sed command.
sed -E '/ORA-01555/!d;:A;N;/\n[A-Za-z]{3} [A-Za-z]{3} [0-9]{2} [0-9]{2}:[0-9]{2}:[0-9]{2} [0-9]{4}$/!bA;:B;N;/\n[A-Za-z]{3} [A-Za-z]{3} [0-9]{2} [0-9]{2}:[0-9]{2}:[0-9]{2} [0-9]{4}$/!bB;s/(.*)\n.*/\1/' infile
If a line contain ORA-01555 --> keep in patern space.
Take the next line in patern space while it's not a date.
Keep the first date in the patern space.
Continue to take the next line while it's not a date.
Delete the last line in the patern space which is a date.
Print the patern space and return to the start to find a new line with ORA-01555.

Answer 2

awk solution:
awk '/ORA-01555/{ f=1 }f && /^[A-Z][a-z]{2} .* [0-9]{4}$/{ date++; if(date>=2) f=date=0 }f' alert.log