diff options
author | unknown <istruewing@chilla.local> | 2007-03-16 10:28:48 +0100 |
---|---|---|
committer | unknown <istruewing@chilla.local> | 2007-03-16 10:28:48 +0100 |
commit | 396f84aa82d371bb0341e7c0c4c37235cca5c9a7 (patch) | |
tree | 88607a0b9f512b37c8ebf05e7d095b79ddcb3c32 /myisam/mi_search.c | |
parent | da9c659c8196d1da63330fc7b1b6710217801a6f (diff) | |
download | mariadb-git-396f84aa82d371bb0341e7c0c4c37235cca5c9a7.tar.gz |
Bug#26231 - select count(*) on myisam table returns wrong value
when index is used
When the table contained TEXT columns with empty contents
('', zero length, but not NULL) _and_ strings starting with
control characters like tabulator or newline, the empty values
were not found in a "records in range" estimate. Hence count(*)
missed these records.
The reason was a different set of search flags used for key
insert and key range estimation.
I decided to fix the set of flags used in range estimation.
Otherwise millions of databases around the world would require
a repair after an upgrade.
The consequence is that the manual must be fixed, which claims
that TEXT columns are compared with "end space padding". This
is true for CHAR/VARCHAR but wrong for TEXT. See also bug 21335.
myisam/mi_range.c:
Bug#26231 - select count(*) on myisam table returns wrong value
when index is used
Added SEARCH_UPDATE to the search flags so that it compares
like write/update/delete operations do. Only so it expects
the keys at the place where they have been inserted.
myisam/mi_search.c:
Bug#26231 - select count(*) on myisam table returns wrong value
when index is used
Added some comments to explain how _mi_get_binary_pack_key()
works.
mysql-test/r/myisam.result:
Bug#26231 - select count(*) on myisam table returns wrong value
when index is used
Added a test.
mysql-test/t/myisam.test:
Bug#26231 - select count(*) on myisam table returns wrong value
when index is used
Added test result.
Diffstat (limited to 'myisam/mi_search.c')
-rw-r--r-- | myisam/mi_search.c | 42 |
1 files changed, 35 insertions, 7 deletions
diff --git a/myisam/mi_search.c b/myisam/mi_search.c index 517fc9c25ff..34cd1ec96fd 100644 --- a/myisam/mi_search.c +++ b/myisam/mi_search.c @@ -918,11 +918,16 @@ uint _mi_get_binary_pack_key(register MI_KEYDEF *keyinfo, uint nod_flag, /* Keys are compressed the following way: - prefix length Packed length of prefix for the prev key. (1 or 3 bytes) + prefix length Packed length of prefix common with prev key (1 or 3 bytes) for each key segment: [is null] Null indicator if can be null (1 byte, zero means null) [length] Packed length if varlength (1 or 3 bytes) + key segment 'length' bytes of key segment value pointer Reference to the data file (last_keyseg->length). + + get_key_length() is a macro. It gets the prefix length from 'page' + and puts it into 'length'. It increments 'page' by 1 or 3, depending + on the packed length of the prefix length. */ get_key_length(length,page); if (length) @@ -935,34 +940,44 @@ uint _mi_get_binary_pack_key(register MI_KEYDEF *keyinfo, uint nod_flag, my_errno=HA_ERR_CRASHED; DBUG_RETURN(0); /* Wrong key */ } - from=key; from_end=key+length; + /* Key is packed against prev key, take prefix from prev key. */ + from= key; + from_end= key + length; } else { - from=page; from_end=page_end; /* Not packed key */ + /* Key is not packed against prev key, take all from page buffer. */ + from= page; + from_end= page_end; } /* - The trouble is that key is split in two parts: - The first part is in from ...from_end-1. - The second part starts at page + The trouble is that key can be split in two parts: + The first part (prefix) is in from .. from_end - 1. + The second part starts at page. + The split can be at every byte position. So we need to check for + the end of the first part before using every byte. */ for (keyseg=keyinfo->seg ; keyseg->type ;keyseg++) { if (keyseg->flag & HA_NULL_PART) { + /* If prefix is used up, switch to rest. */ if (from == from_end) { from=page; from_end=page_end; } if (!(*key++ = *from++)) continue; /* Null part */ } if (keyseg->flag & (HA_VAR_LENGTH | HA_BLOB_PART | HA_SPACE_PACK)) { - /* Get length of dynamic length key part */ + /* If prefix is used up, switch to rest. */ if (from == from_end) { from=page; from_end=page_end; } + /* Get length of dynamic length key part */ if ((length= (*key++ = *from++)) == 255) { + /* If prefix is used up, switch to rest. */ if (from == from_end) { from=page; from_end=page_end; } length= (uint) ((*key++ = *from++)) << 8; + /* If prefix is used up, switch to rest. */ if (from == from_end) { from=page; from_end=page_end; } length+= (uint) ((*key++ = *from++)); } @@ -982,20 +997,33 @@ uint _mi_get_binary_pack_key(register MI_KEYDEF *keyinfo, uint nod_flag, key+=length; from+=length; } + /* + Last segment (type == 0) contains length of data pointer. + If we have mixed key blocks with data pointer and key block pointer, + we have to copy both. + */ length=keyseg->length+nod_flag; if ((tmp=(uint) (from_end-from)) <= length) { + /* Remaining length is less or equal max possible length. */ memcpy(key+tmp,page,length-tmp); /* Get last part of key */ *page_pos= page+length-tmp; } else { + /* + Remaining length is greater than max possible length. + This can happen only if we switched to the new key bytes already. + 'page_end' is calculated with MI_MAX_KEY_BUFF. So it can be far + behind the real end of the key. + */ if (from_end != page_end) { DBUG_PRINT("error",("Error when unpacking key")); my_errno=HA_ERR_CRASHED; DBUG_RETURN(0); /* Error */ } + /* Copy data pointer and, if appropriate, key block pointer. */ memcpy((byte*) key,(byte*) from,(size_t) length); *page_pos= from+length; } |