There have been some requests for detecting precision loss during conversion from float types to integral types. But the current numeric converter just rounds the float number. This proposal provides two classes, that can be used as a Float2IntRounder in boost::numeric::converter. One of them determinates, is there any precision loss, the second class determinates is there precision loss before user supported sign in decimal part. On precision loss an exception boost::numeric::precision_loss_error is thrown. The proposal can be found at https://docs.google.com/document/d/1IsA40cq9kYF31jDM_tyIvVNnvC5-uI9-5ez52pzl... Tests (and usage examples) can be found at https://docs.google.com/document/d/1cbYYDcL1t0jbZ7hUFqXODu3VKoWe4_A5IMdXHN5n... Is there interest in this for Boost? If there is, shall I add documentation?