Content deleted Content added
m Adding short description: "Method for compressing changes over time" |
m linking |
||
Line 10:
If one simply wishes to reconstruct the target given the source and patch, one may simply include the entire target in the patch and "apply" the patch by discarding the source and outputting the target that has been included in the patch; similarly, if the source and target have the same size one may create a simple patch by [[XOR]]ing source and target. In both these cases, the patch will be as large as the target. As these examples show, if the only concern is reconstruction of target, this is easily done, at the expense of a large patch, and the main concern for general-purpose binary differencing is reducing the patch size.
For structured data especially, one has other concerns, which largely fall under "usability" – for example, if one is [[file comparison|comparing]] two documents, one generally wishes to know ''which'' sections have changed, or if some sections have been moved around – one wishes to understand ''how'' the documents differ. For instance "here 'cat' was changed to 'dog', and paragraph 13 was moved to paragraph 14". One may also wish to have ''robust'' differences – for example, if two documents A and B differ in paragraph 13, one may wish to be able to apply this patch even if one has changed paragraph 7 of A. An example of this is in diff, which shows which lines changed, and where the context format allows robustness and improves human readability.
Other concerns include computational efficiency, as for data compression – finding a small patch can be very time and memory intensive.
| |||