HAPI-UR is a method for inferring haplotypes in large datasets of unrelated and/or trio or duo individuals. The implementation, source code, and documentation for HAPI-UR are available here. Details of the method are in the paper Phasing of many thousands of genotyped samples.
HAPI is a method for inferring haplotypes in nuclear families that is orders of magnitude faster than other pedigree-based phasing methods. The software and documentation are available here. The algorithm is described in the paper "Rapid Haplotype Inference for Nuclear Families". While the problem of inferring haplotypes in family data has been shown to be NP-hard, in practice HAPI has polynomial runtime. This is possible because real genotype data exhibits a relatively small number of recombinations, with few children ever exhibiting recombination at the same marker position.
I developed an extension to HapMix for inferring local ancestry in multi-way admixed populations. This method was used as part of a composite method to produce local ancestry results for the 1,000 Genomes Project. Please contact me if you are interested in using this method.