We present photometric and spectroscopic observations of SN 2007if, an overluminous (M V = −20.4), red (B − V = 0.16 at B-band maximum), slow-rising (t rise = 24 days) type Ia supernova (SN Ia) in a very faint (M g = −14.10) host galaxy. A spectrum at 5 days past B-band maximum light is a direct match to the super-Chandrasekhar-mass candidate SN Ia 2003fg, showing Si II and C II at ∼ 9000 km s −1 . A high signal-to-noise co-addition of the SN spectral time series reveals no Na I D absorption, suggesting negligible reddening in the host galaxy, and the late-time color evolution has the same slope as the Lira relation for normal SNe Ia. The ejecta appear to be well mixed, with no strong maximum in I-band and a diversity of iron-peak lines appearing in near-maximum-light spectra. SN 2007if also displays a plateau in the Si II velocity extending as late as +10 days, which we interpret as evidence for an overdense shell in the SN ejecta. We calculate the bolometric light curve of the SN and use it and the Si II velocity evolution to constrain the mass of the shell and the underlying SN ejecta, and demonstrate that SN 2007if is strongly inconsistent with a Chandrasekhar-mass scenario. Within the context of a "tamped detonation" model appropriate for double-degenerate mergers, and assuming no host extinction, we estimate the total mass of the system to be 2.4 ± 0.2 M ⊙ , with 1.6 ± 0.1 M ⊙ of 56 Ni and with 0.3-0.5 M ⊙ in the form of an envelope of unburned carbon/oxygen. Our modeling demonstrates that the kinematics of shell entrainment provide a more efficient mechanism than incomplete nuclear burning for producing the low velocities typical of super-Chandrasekhar-mass SNe Ia.