Scroll to navigation

PMMERGELABELS(3) Library Functions Manual PMMERGELABELS(3)


pmMergeLabels, pmMergeLabelSets - merge sets of performance metric labels


#include <pcp/pmapi.h>

int pmMergeLabels(char **sets, int nsets, char *buffer, int length);

int pmMergeLabelSets(pmLabelSet **sets, int nsets, char *buffer, int length, int (*filter)(const pmLabel *, const char *, void *), void *arg);

cc ... -lpcp


from pcp import pmapi

buffer = pmapi.pmContext().pmMergeLabels(sets)

buffer = pmapi.pmContext().pmMergeLabelSets(sets, filter)


pmMergeLabels takes multiple (nsets) performance metric label sets and merges them into a single result buffer of length bytes. Both the input sets and the result buffer are name:value pairs in the "JSONB" format described on pmLookupLabels(3).

The pmMergeLabelSets interface serves the same purpose, but allows for indexed sets of labels to be merged. The format of the pmLabelSet data structure is described in detail in pmLookupLabels(3).

Although names may repeat across the provided label sets, duplicate names are not allowed in the final buffer. Any label names occuring in more than one of the input label sets are reduced to one using the rules described in the "PRECEDENCE" section of pmLookupLabels. The position of each element in the sets array is significant in terms of the precedence rules - earlier positions are taken to be of lower precedence to later positions.

Values must be primitive JSON entities (e.g. numbers, strings), one-dimensional arrays or maps (i.e. simple associative arrays).

In addition to using indexed label sets the pmMergeLabelSets interface provides an optional filter callback function. If non-NULL, this function will be called for each label that would be added to the output buffer, allowing finer-grained control over the final merged set. This mechanism can be used to filter individual labels based on their name, value, and/or flags. If the filter function returns zero (false), then the given label is filtered from the resulting set. Any non-zero return value indicates that the label should be included in the buffer.


import sys
import json
from pcp import pmapi
import cpmapi as c_api
def merge_callback(label, jsondata, data=None):

d = json.loads(jsondata)
return 0 ctx = pmapi.pmContext() for metric in sys.argv[1:]:
pmid = ctx.pmLookupName(metric)[0]
lset = ctx.pmLookupLabels(pmid)
labelsD = {}
ctx.pmMergeLabelSets(lset, merge_callback)
print("== %s ===" % metric)
for n,v in labelsD.items():
print(" %s = %s" % (n,v))


On success, both pmMergeLabels and pmMergeLabelSets returns the number of bytes written into the supplied buffer.

Failure to parse the input strings, failure to allocate memory, or any internal inconsistencies found will result in a negative return code.


pminfo(1), PMAPI(3) and pmLookupLabels(3).

PCP Performance Co-Pilot