在Linux中如何检查连续模式中是否有2种不同的模式

Question

I've an ASCII text file which I'm validating. The file contains contexts of 2 types:

Necessary Context: One which should be present at least once in its exact position.
Optional Context: One which may or may not be present, but if present should hold its proper place.

The detail look of the file:

[INDEX] <-- optional context, but if present should be the first context immediately followed by [FEATURE_ID], else file invalid
[FEATURE_ID] <-- necessary context and should always be immediately followed by [FEATURE_REV], else file is invalid. If [INDEX] context there then this should be the second CONTEXT in file else first.
[FEATURE_REV] <-- necessary context (must exist one per FEATURE_ID) and should always be immediately after [FEATURE_ID], else file is invalid.
[PRL_ID] <-- optional context, but if present should always be immediately after [FEATURE_REV], else file invalid
[NO_OF_BYTES] <--optional context, but if present, should always be immediately after [PRL_ID] if it is present, else immediately after [FEATURE_REV] if [PRL_ID] not present. Otherwise file invalid.
[NO_OF_SIGNIF_BITS] <-- optional context, but if present should always be between [NO_OF_BYTES] ( can be only present if [NO_OF_BYTES] present else not) and [CRC], else file invalid
[CRC] <-- necessary context,(must exist one per FEATURE_ID and FEATURE_REV). This is always the last context.

Note, there might be multiple [FEATURE_ID] contexts in a valid File, and in all cases the other contexts leading and following it, should follow same place holding rule. Something like this:

Validfile_1:

[FEATURE_ID]
[FEATURE_REV]
[CRC]

[INDEX]
[FEATURE_ID]
[FEATURE_REV]
[CRC]

Validfile_2:

[FEATURE_ID]
[FEATURE_REV]
[NO_OF_BYTES]
[CRC]

[INDEX]
[FEATURE_ID]
[FEATURE_REV]
[PRL_ID]
[NO_OF_BYTES]
[NO_OF_SIGNIF_BITS]
[CRC]

Validfile_3

[FEATURE_ID]
[FEATURE_REV]
[CRC]

Invalidfile_1 (order of contexts not ok):

[FEATURE_ID]
[INDEX]
[FEATURE_REV]
[NO_OF_BYTES]
[CRC]
[PRL_ID]

Invalidfile_2(FEATURE_REV or CRC can never exist without a FEATURE_ID):

[FEATURE_REV]
[NO_OF_BYTES]
[CRC]

Invalidfile_3 ( NO_OF_SIGNIF_BITS cannot exist without NO_OF_BYTES)

[FEATURE_ID]
[FEATURE_REV]
[NO_OF_SIGNIF_BITS]
[CRC]

I'm trying to achieve this in a linux script via multiple if else statements and egreps , but the lines of code are becoming more and complex.

The code that I'm going for:

f_id_c=`egrep "[ ]*\[FEATURE_ID=[0-9].*\][ ]*" $1 | wc -l`
f_rev_c=`egrep "[ ]*\[FEATURE_REV=[0-9].*\][ ]*" $1 | wc -l`
crc_c=`egrep "[ ]*\[CRC\][ ]*" $1 | wc -l`
[[ $((f_id_c)) -eq 0 ]] && { echo "Invalid! No [FEATURE_ID=] context defined in profile file !"; exit 1; }
[[ $((f_rev_c)) -ne $((f_id_c)) ]] && { echo "Invalid! Not all [FEATURE_REV=] contexts have leading [FEATURE_ID=] defined"; exit 1; }
[[ $((crc_c)) -ne $((f_id_c)) ]] && { echo "Invalid! Not all [CRC] contexts have leading [FEATURE_ID=] defined"; exit 1; }
for (i=0;i<f_id_c;i++)
  do
    // Have a check with SED that will confirm there is a [FEATURE_REV=] immediately following [FEATURE_ID=]
  done

Can someone suggest a compact awk script or sed manipulation where I can achieve all the above validation.

Answer 1

You're going to want a FSM something like this:

$ cat tst.awk
BEGIN {
    # define the allowed state transitions
    ns["IDLE","INDEX"]
    ns["IDLE","FEATURE_ID"]
    ns["INDEX","FEATURE_ID"]
    ns["FEATURE_ID","FEATURE_REV"]
    ns["FEATURE_REV","PRL_ID"]
    ns["FEATURE_REV","NO_OF_BYTES"]
    ns["FEATURE_REV","CRC"]
    ns["PRL_ID","NO_OF_BYTES"]
    ns["PRL_ID","CRC"]
    ns["NO_OF_BYTES","NO_OF_SIGNIF_BITS"]
    ns["NO_OF_BYTES","CRC"]
    ns["NO_OF_SIGNIF_BITS","CRC"]
    ns["CRC","INDEX"]
    ns["CRC","FEATURE_ID"]

    # create a regexp of the state names for use in match()
    for (state in ns) {
        sub(SUBSEP".*","",state)
        if (!seen[state]++) {
            states = states (states ? "|" : "") state
        }
    }

    # set the initial state
    state = "IDLE"
}

# parse the input
match($0,states) {
    nextState = substr($0,RSTART,RLENGTH)
    if ( ! ((state,nextState) in ns) ) {
        print "ERROR", NR, state, nextState, $0 | "cat>&2"
        exit 1
    }
    state = nextState
}

When run against your posted sample input file:

$ cat file
....
[FEATURE_ID]
[FEATURE_REV]
...
...
[CRC]

[INDEX]
[FEATURE_ID]
[FEATURE_REV]
...
...
...
[CRC]
$
$ awk -f tst.awk file
$

it produces no output, as expected since the sample you provided contains no errors for it to find.