Wrong result when paring some diff that contains deleted sql comment starting with --

[jiashanyao]

Here is a test diff input to reproduce the bug

diff --git a/test.sql b/test.sql
index 305feaa..70865e7 100644
--- a/test.sql
+++ b/test.sql
@@ -1,10 +1,9 @@
--- Person
-create table person
-( id                   		bigint,
-  primary key ( id ) );
-
-
 -- Photo
 create table photo
 ( id                      bigint,
   primary key (id ) );
+
+-- Person
+create table person
+( id                   		bigint,
+  primary key ( id ) );

And the parsed result is

[
  {
    "chunks": [
      {
        "content": "@@ -1,10 +1,9 @@",
        "changes": [
          {
            "type": "del",
            "del": true,
            "ln": 1,
            "content": "-create table person"
          },
          {
            "type": "del",
            "del": true,
            "ln": 2,
            "content": "-( id                   \t\tbigint,"
          },
          {
            "type": "del",
            "del": true,
            "ln": 3,
            "content": "-  primary key ( id ) );"
          },
          {
            "type": "del",
            "del": true,
            "ln": 4,
            "content": "-"
          },
          {
            "type": "del",
            "del": true,
            "ln": 5,
            "content": "-"
          },
          {
            "type": "normal",
            "normal": true,
            "ln1": 6,
            "ln2": 1,
            "content": " -- Photo"
          },
          {
            "type": "normal",
            "normal": true,
            "ln1": 7,
            "ln2": 2,
            "content": " create table photo"
          },
          {
            "type": "normal",
            "normal": true,
            "ln1": 8,
            "ln2": 3,
            "content": " ( id                      bigint,"
          },
          {
            "type": "normal",
            "normal": true,
            "ln1": 9,
            "ln2": 4,
            "content": "   primary key (id ) );"
          },
          {
            "type": "add",
            "add": true,
            "ln": 5,
            "content": "+"
          },
          {
            "type": "add",
            "add": true,
            "ln": 6,
            "content": "+-- Person"
          },
          {
            "type": "add",
            "add": true,
            "ln": 7,
            "content": "+create table person"
          },
          {
            "type": "add",
            "add": true,
            "ln": 8,
            "content": "+( id                   \t\tbigint,"
          },
          {
            "type": "add",
            "add": true,
            "ln": 9,
            "content": "+  primary key ( id ) );"
          }
        ],
        "oldStart": 1,
        "oldLines": 10,
        "newStart": 1,
        "newLines": 9
      }
    ],
    "deletions": 0,
    "additions": 0,
    "from": "test.sql",
    "to": "test.sql",
    "index": [
      "305feaa..70865e7",
      "100644"
    ]
  },
  {
    "chunks": [],
    "deletions": 5,
    "additions": 5,
    "from": "Person"
  }
]

which is wrong.

[mironiasty]

I found similar issue, but for files that contains lines like ++ some text. It looks like it could be fixed in two ways:

• By improving regex to better detect added/removed files from patch
• By checking if toFile line is right after fromFile

@sergeyt If I would create correct fix for it, would it be merged? (AKA is this repo still maintained)

[sergeyt]

@mironiasty Sure. As usual, fixes should be covered by unit tests and all tests should be passing. I am happy to release new version

[mironiasty]

small update: I was working on fix. But looks like this bug can't be really fixed without using information from chunk, to get number of lines with changed data. Any other solution I tried, either won't really fix it or it will introduce some breaking changes. I hope I will have some time soon to implement proper fix.