We present a new data structure for Huffman coding in which in addition to sending symbols in order of their appearance in the Huffman tree one needs to send codes of all circular leaf nodes (nodes with two adjacent external nodes), the number of which is always bounded above by half the number of symbols. We decode the text by using the memory efficient data structure proposed by Chen et al. [Inform. Process. Lett. 69 (1999) 119
Rezaul Alam Chowdhury, M. Kaykobad, Irwin King