------------------------------------------------------------------------------- Alignment thresholds: Alignment threshold are used to prevent aligning of certain text. Consider this example when aligning assignments: a = 2; gld.settings[somevariable] = 5; Without thresholds, this would result in: a = 2; gld.settings[somevariable] = 5; However, this is likely not desired. So to handle this, we set a limit on the number of columns that a alignment may move a token. If we set the threshold to, say, 3 characters, then the two would not be aligned together because 'a = 2' would have to be moved by 25 columns. Here's a more complicated case: a = 1; bb = 2; ccc = 3; dddd = 4; eeeee = 5; ffffff = 6; In this case, we may want this output: a = 1; bb = 2; ccc = 3; dddd = 4; eeeee = 5; ffffff = 6; But, with a threshold of 3, 'a' and 'ffffff' cannot be aligned together. So how should these thresholds be implemented? One approach is to use the threshold as the maximum difference between the current maximum and the new maximum. Using this approach, all 6 lines above are aligned, because the change for each is 1 column. Consider this with a line-span of 3 and a column-threshold of 3. Note that we'll now need to keep track of the last line processed. (line and column numbers added for clarity) 111111 123456789012345 --------------- 1 | a = 1; 2 | eeeee = 5; 3 | ffffff = 6; On line 1, the maxcol is set to 3. On line 2, the maxcol would be set to 7, except that exceeds the threshold. On line 3, the maxcol would be set to 8, except that exceeds the threshold. So, it ends up with one item in the align stack - line 1. Then we pick up from where we left off - on line 2. On line 2, the maxcol is set to 7. On line 3, the maxcol is set to 8. End result: 111111 123456789012345 --------------- 1 | a = 1; 2 | eeeee = 5; 3 | ffffff = 6; Now lets get tricky. 111111 123456789012345 --------------- 1 | a = 1; 2 | eeeee = 5; 3 | ccc = 3; 4 | ffffff = 6; How should this be aligned? I think we need to add another list - a 'skipped' list, that runs in parallel with the 'aligned' list. On line 1, the maxcol is set to 3, added to the aligned list. On line 2, col exceeds the threshold, so it is added to the skipped list. On line 3, the maxcol is set to 5, added to the aligned list. Now, after an item is added to the aligned list, the skipped list must be re-scanned. We now see that line 2 is only 2 columns different, so it is added and maxcol = 7. The skipped list is now empty. On line 4, the maxcol is set to 8, added to the aligned list. So the output is: 111111 123456789012345 --------------- 1 | a = 1; 2 | eeeee = 5; 3 | ccc = 3; 4 | ffffff = 6; Now for a case where the skipped list is not absorbed: 111111 123456789012345 --------------- 1 | a = 1; 2 | iiiiiiiiiiiiieeeee = 5; 3 | ccc = 3; 4 | ffffff = 6; On line 1, the maxcol is set to 3, added to the aligned list. On line 2, col exceeds the threshold, so it is added to the skipped list. On line 3, the maxcol is set to 5, added to the aligned list. Skipped list is scanned, nothing moved from the skipped list to the aligned list. On line 4, the maxcol is set to 8, added to the aligned list. So the output is: 111111 123456789012345 --------------- 1 | a = 1; 2 | iiiiiiiiiiiiieeeee = 5; 3 | ccc = 3; 4 | ffffff = 6; As a reminder, the old system would have produced: 1 | a = 1; 2 | iiiiiiiiiiiiieeeee = 5; 3 | ccc = 3; 4 | ffffff = 6; Which is probably not wanted. --===---===---===---===-- Absolute thresholds To get a better grip on what thresholds do the absolute thresholds were introduced. An absolute threshold means that the item, in this case the assign statement, is never moved more than the threshold value. See the below example: Relative threshold = 10 000000000111111111122222222223 123456789012345678901234567890 1| a = 1 2| aaaaa = 1 3| bbbbbbbbbbb = 2 4| ccccccccccccccccccccc = 3 Absolute threshold: 000000000111111111122222222223 123456789012345678901234567890 1| a = 1 2| aaaaa = 1 3| bbbbbbbbbbb = 2 4| cccccccccccccccccccc = 3 --===---===---===---===-- How to do this generically in the code... Easy case - one item per line. Here's the functions and what each do: align_start(line_span, col_thresh) align_add(pc) align_flush() align_newline(count) align_end() For each entry, a sequence number is kept. align_start(line_span, col_thresh) - initializes the align and skipped lists - zero max_column - zero cur_seqnum - zero nl_count align_add(pc) - update cur_seqnum, assign to item - if item column is within threshold - zero nl_count - add to the aligned list - if item column is > max_col - update max_col - re-adds all items on the skipped list - if item column is not within threshold, add to skipped list align_newline(count) - adds count to nl_count - if nl_count > line_span, call align_flush() align_flush() - step through all the items in aligned list and align the items to max_column - keep the seq_num of the last aligned item - zero max_column - remove all items with a seq_num < last aligned seq_num - call align_add on all remaining items in the skipped list align_end() - call align_flush - clear the lists - free resources, etc Example usage: Aligning trailing comments void align_trailing_comments(void) { chunk_t *pc; align_start(cpd.settings[UO_align_right_cmt_span], cpd.settings[UO_align_right_cmt_thresh]); for(pc = chunk_get_head(); pc != NULL; pc = chunk_get_next(pc)) { if ((pc->flags & PCF_RIGHT_COMMENT) != 0) { align_add(pc); } else if (chunk_is_newline(pc)) { align_newline(pc->nl_count); } } align_end(); }