add initial files

This commit is contained in:
wangziao 2025-09-20 19:26:03 -07:00
parent 89ae608cf8
commit a6541962e5
3 changed files with 132 additions and 2 deletions

View File

@ -1,3 +1,10 @@
# lcs-diff
# Toy LCS-based Diff program
Toy Longest-Common-Subsequence based diff and patch program, minimizing the edit script's sum of line-addition and line-deletions.
Toy Longest-Common-Subsequence based diff and patch program, minimizing the edit script's sum of line-addition and line-deletions.
Example usage:
```bash
python diff.py a1 b1 > patch1
python patch.py a1 patch1 > c1 # patch is not in-place
python diff.py a1 c1 # verify that a1 and c1 has the same content
```

58
diff.py Normal file
View File

@ -0,0 +1,58 @@
import argparse
parser = argparse.ArgumentParser(description="Compare two documents line by line")
parser.add_argument('a', help="reference document")
parser.add_argument('b', help="target document")
args = parser.parse_args()
fa = open(args.a,'r')
la = fa.read().split('\n')
fb = open(args.b,'r')
lb = fb.read().split('\n')
# Solve the Longest Common Subsequence (LCS) subproblem with dynamic programming
dp = [[0] * (len(lb)+1) for _ in range(len(la)+1)]
# dp[i+1][j+1] stores the LCS length between la[i] and lb[j]
for i in range(len(la)):
for j in range(len(lb)):
if la[i]==lb[j]:
dp[i+1][j+1] = dp[i][j] + 1
else:
dp[i+1][j+1] = max(dp[i][j+1], dp[i+1][j])
# Backtrack to find one scheme to reduce both a and b to their LCS
diffs = []
i,j = len(la)-1, len(lb)-1
while i>=0 or j>=0: # current position (i+1, j+1), try to move to (0, 0)
oi, oj = i, j
while j >= 0 and dp[i+1][j+1]==dp[i+1][j]: # can safely delete lb[j]
j -= 1
while i >= 0 and dp[i+1][j+1]==dp[i][j+1]: # can safely delete la[i]
i -= 1
if i==oi and j==oj:
assert dp[i+1][j+1]==dp[i][j]+1 and la[i]==lb[j]
i,j = i-1,j-1
else:
diffs.append((oi,i,oj,j))
diffs.reverse()
def describe(oi,i,oj,j):
# by diff convension line numbering starts from 1
oi,i,oj,j = oi+1,i+1,oj+1,j+1
def intv(x,y): # simplify expression if interval is one line
return str(x) if x==y else f"{x},{y}"
if i==oi:
print(f"{i}a" + intv(j+1,oj))
elif j==oj:
print(intv(i+1,oi) + f"d{j}")
else:
print(intv(i+1,oi) + "c" + intv(j+1,oj))
for oi,i,oj,j in diffs:
# delete (i,oi], add (j,oj]
describe(oi,i,oj,j)
for p in range(i+1,oi+1):
print("< " + la[p])
if oi>i and oj>j:
print("---")
for p in range(j+1,oj+1):
print("> " + lb[p])

65
patch.py Normal file
View File

@ -0,0 +1,65 @@
import argparse
parser = argparse.ArgumentParser(description="Prints to stdout the patched document")
parser.add_argument('original', help='original document')
parser.add_argument('patch', help='patch file')
args = parser.parse_args()
fa = open(args.original,'r')
la = fa.read().split('\n')
# a workspace for us to edit, each element being a list that 1) originally contains a corresponding line
# in a, 2) may be emptied if that line is removed, 3) may be where we insert new content
ws = [[l] for l in la]
fp = open(args.patch,'r')
lp = fp.read().split('\n')
def str2intv(s):
if "," in s:
x, y = s.split(',')
else:
x, y = s, s
return int(x)-1, int(y)
def ws_rm(desc, lp, nb):
l,r = str2intv(desc)
for p in range(l,r):
nb += 1
assert lp[nb][:2] == "< "
assert ws[p][0] == lp[nb][2:],\
f"Patch file wants to delete '{lp[nb][2:]}', but originally line {p} is '{ws[p][0]}'"
ws[p] = list()
return nb + 1, l
def ws_insert(p, desc, lp, nb):
l,r = str2intv(desc)
for _ in range(l,r):
nb += 1
assert lp[nb][:2] == "> "
ws[p].append(lp[nb][2:])
return nb + 1
nb = 0
while nb < len(lp):
line = lp[nb]
if line:
assert line[0] != '<' and line[0] != '>', f"Expected description at line {nb}, but got {line}"
else:
nb += 1
continue
idx_d = line.find('d')
if idx_d != -1:
nb, _ = ws_rm(line[:idx_d], lp, nb)
continue
idx_a = line.find('a')
if idx_a != -1:
p = int(line[:idx_a]) - 1
nb = ws_insert(p, line[idx_a+1:], lp, nb)
continue
idx_c = line.find('c')
if idx_c != -1:
nb, p = ws_rm(line[:idx_c], lp, nb)
nb = ws_insert(p, line[idx_c+1:], lp, nb)
continue
assert False, f"Unreachable. Cannot parse line {nb}: {line}"
flatws = [line for _ in ws for line in _]
print('\n'.join(flatws), end="") # Don't want to print additional new line