Links in a practical network may have different functions, which makes the original network a combination of some functional subnetworks. Here, by a model of coupled oscillators, we investigate how such functional subnetworks are evolved and developed according to the network structure and dynamics. In particular, we study the case of evolutionary clustered networks in which the function of each link (either attractive or repulsive coupling) is updated by the local dynamics. It is found that, during the process of system evolution, the network is gradually stabilized into a particular form in which the attractive (repulsive) subnetwork consists only the intralinks (interlinks). Based on the properties of subnetwork evolution, we also propose a new algorithm for network partition which is distinguished by the convenient operation and fast computing speed. The past decade has witnessed the blooming of network science, in which one important issue is to explore the interplay between the network structure and dynamics [1,2]. While the influences of the network structure on dynamics have been intensively studied in the past [3], recently attentions have also been paid to the influences of the network dynamics on structure, i.e. the evolution of complex networks driven by dynamics [4,5,6,7,8,9,10]. In Ref.[5] it has been shown that, rewiring network links according to the node synchronization, a random network can be gradually developed to a smallworld network. In Ref. [7] it has been shown that, driven by node synchronization, the weight of the network links can be developed to a particular form in favor of global network synchronization. Besides network evolution, dynamics has been also used for network detection, e.g., detecting the modular structures in clustered complex networks [6,7,8,9,10,11].It has been well recognized that links in a practical network are usually different from each other. In previous studies, this has been mainly reflected in the variation of the weight of the network links, i.e., the weighted network [1]. Weighted network, however, describes only the case of single-function networks, i.e. all links in the network have the same function, but failing to describe the situation of multi-function networks in which the network links have the diverse functions. A type of commonly seen multi-function networks in practice is the cooperation-competition network (CCN) [12,13], in which the network links are divided into two groups of opposite functions. For instance, in the nervous network of the human brain, the synapses are roughly divided into two groups, excitatory and inhibitory, which play the contrary roles to the neuron activities [12]. Another typical example of CCN is the relationship network shown in the prisoner's dilemma game, in which each suspect may either cooperate with (remain silent) or defect from (betray) the other suspects [13]. * Corresponding author. Email address: wangxg@zju.edu.cn For multi-function networks like CCN, to facilitate the analysis, it will be convenient if we tr...