Support DELIMITER in batch input

changelog.md

@@ -4,6 +4,7 @@ Upcoming (TBD)
 Bug Fixes
 --------
 * Open the audit log and the `\T`/`\o` output files as UTF-8, so non-ASCII output no longer fails under non-UTF-8 locales.
+* Support `DELIMITER` in batch input (`mycli < file.sql`, `--batch`), so dumps with triggers or routines restore.
 
 
 v2.30.4 (2026/11/12)

mycli/AUTHORS.txt

@@ -123,6 +123,7 @@ Contributors:
   * Wei-Lun Li
   * Shaurya Kesarwani
   * Alex Castillo
+  * Jay Hemnani
 
 
 Created by:

mycli/packages/utils/batch_utils.py

@@ -1,25 +1,52 @@
+import re
 from typing import IO, Generator
 
 import sqlglot
 import sqlparse
 
 MAX_MULTILINE_BATCH_STATEMENT = 5000
 
+DELIMITER_RE = re.compile(r'^\s*delimiter\s+(\S+)\s*$', re.IGNORECASE)
+
+
+def _has_tokens(text: str) -> bool:
+    try:
+        return bool(sqlglot.tokenize(text, read='mysql'))
+    except sqlglot.errors.TokenError:
+        return True
+
 
 def statements_from_filehandle(file_h: IO) -> Generator[tuple[str, int], None, None]:
     statements = ''
     line_counter = 0
     batch_counter = 0
+    delimiter = ';'
     for batch_text in file_h:
         line_counter += 1
         if line_counter > MAX_MULTILINE_BATCH_STATEMENT:
             raise ValueError(f'Saw single input statement greater than {MAX_MULTILINE_BATCH_STATEMENT} lines; assuming a parsing error.')
+        # DELIMITER is a client command, as in the mysql client: switch delimiters, don't send it
+        delimiter_match = DELIMITER_RE.match(batch_text)
+        if delimiter_match and not _has_tokens(statements):
+            delimiter = delimiter_match.group(1)
+            statements = ''
+            line_counter = 0
+            continue
         statements += batch_text
+        if delimiter != ';':
+            statement = statements.rstrip()
+            # A custom delimiter is only recognized at the end of a line, so
+            # two statements on one line are sent together as one
+            if statement.endswith(delimiter):
+                yield (statement[: -len(delimiter)].strip(), batch_counter)
+                batch_counter += 1
+                statements = ''
+                line_counter = 0
+            continue
         try:
             tokens = sqlglot.tokenize(statements, read='mysql')
             if not tokens:
                 continue
-            # we don't yet handle changing the delimiter within the batch input
             if tokens[-1].text == ';':
                 # The advantage of sqlparse for splitting is that it preserves the input.
                 # https://github.com/tobymao/sqlglot/issues/2587#issuecomment-1823109501

mycli_test/pytests/test_batch_utils.py

@@ -99,3 +99,12 @@ def fake_tokenize(sql: str, read: str):
         ('select 1;', 0),
         ('select 2;', 1),
     ]
+
+
+def test_statements_from_filehandle_handles_delimiter_blocks() -> None:
+    statements = collect_statements('DELIMITER //\nCREATE PROCEDURE p1()\nBEGIN\n  SELECT 1;\nEND //\nDELIMITER ;\nCALL p1();\n')
+
+    assert statements == [
+        ('CREATE PROCEDURE p1()\nBEGIN\n  SELECT 1;\nEND', 0),
+        ('CALL p1();', 1),
+    ]