Question

I’m making a program to calculate latency from a tcpdump/pcap file and I want to be able to specify rules on the command line to correlate packets — i.e. find the time taken between sending a packet matching rule A to receiving a packet matching rule B (concrete example would be a FIX NewOrderSingle being sent and a corresponding FIX ExecutionReport being received).

This is an example of the fields in the packet (before they’ve been converted into dictionary form) — I’m testing the numerical version of the field (in parentheses) rather than the English version:

    BeginString (8): FIX.4.2
    BodyLength (9): 132
    MsgType (35): D (ORDER SINGLE)
    SenderCompID (49): XXXX
    TargetCompID (56): EXCHANGE
    MsgSeqNum (34): 1409104
    SendingTime (52): 20100723-12:49:52.296
    Side (54): 1 (BUY)
    Symbol (55): A002
    ClOrdID (11): BUY704552
    OrderQty (38): 1000
    OrdType (40): 2 (LIMIT)
    Price (44): 130002
    TimeInForce (59): 3 (IMMEDIATE OR CANCEL)
    QuoteID (117): A002
    RelatdSym (46): A002
    CheckSum (10): 219 [correct]

Currently I have the arguments coming off the command line into a nested list:

[[35, 'D'], [55, 'A002']]

(where the first element of each sublist is the field number and second is the value)

I’ve tried iterating over this list of rules to accumulate a lambda expression:

for field, value in args.send["fields_filter"]:
    if matchers["send"] == None:
        matchers["send"] = lambda fix : field in fix and fix[field] == value
    else:
        matchers["send"] = lambda fix : field in fix and fix[field] == value and matchers["send"](fix)

When I run the program though, I get the output:

RuntimeError: maximum recursion depth exceeded in cmp

Lambdas are late-binding? So does this apply to all identifiers in the expression or just those passed in as arguments? It seems the former is true

What’s the best way to achieve this functionality? I feel like I’m going about this the wrong way currently. Maybe this is a bad use of lambda expressions, but I don’t know a better alternative for this.

Answer 1

Don’t use lambdas. They are late binding. Perhaps you want a partial from functools, but even that seems too complex.

Your data coming in has field names, numbers and values, right?

Your command-line parameters use field numbers and values, right?

You want a dictionary keyed by field number. In that case, you don’t need any complex lookups. You just want something like this.

def match( packet_dict, criteria_list ):
    t = [ packet_dict[f] == v for f,v in criteria_list ]
    return any( t )

Something like that should handle everything for you.