Full_Name: Karl Waclawek Version: LMDB 0.9.22 OS: Windows 10 URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (2607:fea8:7a80:814:3ca9:e8f6:801c:d6f8) The documentation in lmdb.h for MDB_GET_MULTIPLE, MDB_NEXT_MULTIPLE and MDB_PREV_MULTIPLE seems to have two issues: 1) The documentation indicates that both, the key and data, are returned, but in reality only the data is returned. 2) The documentation states that MDB_GET_MULTIPLE moves the cursor. This does not seem to be true.
> Full_Name: Karl Waclawek > Version: LMDB 0.9.22 > OS: Windows 10 > URL: ftp://ftp.openldap.org/incoming/ > Submission from: (NULL) (2607:fea8:7a80:814:3ca9:e8f6:801c:d6f8) > > > The documentation in lmdb.h for MDB_GET_MULTIPLE, MDB_NEXT_MULTIPLE and > MDB_PREV_MULTIPLE seems to have two issues: > > 1) The documentation indicates that both, the key and data, are returned, but in > reality only the data is returned. Looks like you're right. Since these are DUPs the key would be the same every time so there's no need to return it on each call. Doc fixed in git mdb.master. > 2) The documentation states that MDB_GET_MULTIPLE moves the cursor. This does > not seem to be true. The doc is correct here. The cursor is advanced far enough that a MDB_NEXT_MULTIPLE will return the following data, as opposed to just returning the same data again.
changed notes changed state Open to Test moved from Incoming to Documentation
As far as point 2 is concerned I was basing it on this piece of code, where it looks like MDB_GET_MULTIPLE just checks some flags and then goes to the fetchm label where the code does nothing but retrieve data: case MDB_GET_MULTIPLE: if (data == NULL || !(mc->mc_flags & C_INITIALIZED)) { rc = EINVAL; break; } if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { rc = MDB_INCOMPATIBLE; break; } rc = MDB_SUCCESS; if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) || (mc->mc_xcursor->mx_cursor.mc_flags & C_EOF)) break; goto fetchm; case MDB_NEXT_MULTIPLE: if (data == NULL) { rc = EINVAL; break; } if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { rc = MDB_INCOMPATIBLE; break; } rc = mdb_cursor_next(mc, key, data, MDB_NEXT_DUP); if (rc == MDB_SUCCESS) { if (mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { MDB_cursor *mx; fetchm: mx = &mc->mc_xcursor->mx_cursor; data->mv_size = NUMKEYS(mx->mc_pg[mx->mc_top]) * mx->mc_db->md_pad; data->mv_data = METADATA(mx->mc_pg[mx->mc_top]); mx->mc_ki[mx->mc_top] = NUMKEYS(mx->mc_pg[mx->mc_top])-1; } else { rc = MDB_NOTFOUND; } } break; On Tue, Aug 28, 2018 at 8:26 PM Howard Chu <openldap-its@openldap.org> wrote: > > Full_Name: Karl Waclawek > > Version: LMDB 0.9.22 > > OS: Windows 10 > > URL: ftp://ftp.openldap.org/incoming/ > > Submission from: (NULL) (2607:fea8:7a80:814:3ca9:e8f6:801c:d6f8) > > > > > > The documentation in lmdb.h for MDB_GET_MULTIPLE, MDB_NEXT_MULTIPLE and > > MDB_PREV_MULTIPLE seems to have two issues: > > > > 1) The documentation indicates that both, the key and data, are > returned, but > in > > reality only the data is returned. > > Looks like you're right. Since these are DUPs the key would be the same > every time so there's no need to return it on each call. > Doc fixed in git mdb.master. > > > 2) The documentation states that MDB_GET_MULTIPLE moves the cursor. This > does > > not seem to be true. > > The doc is correct here. The cursor is advanced far enough that a > MDB_NEXT_MULTIPLE will return the following data, as opposed to just > returning the same data again. >
karl@waclawek.net wrote: > --00000000000018b43a057492d87f > Content-Type: text/plain; charset="UTF-8" > > As far as point 2 is concerned I was basing it on this piece of code, where > it looks like MDB_GET_MULTIPLE just checks some flags and then goes to the > fetchm label where the code does nothing but retrieve data: You're not reading the code correctly. See below. > > case MDB_GET_MULTIPLE: > if (data == NULL || !(mc->mc_flags & C_INITIALIZED)) { > rc = EINVAL; > break; > } > if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { > rc = MDB_INCOMPATIBLE; > break; > } > rc = MDB_SUCCESS; > if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) || > (mc->mc_xcursor->mx_cursor.mc_flags & C_EOF)) > break; > goto fetchm; > case MDB_NEXT_MULTIPLE: > if (data == NULL) { > rc = EINVAL; > break; > } > if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { > rc = MDB_INCOMPATIBLE; > break; > } > rc = mdb_cursor_next(mc, key, data, MDB_NEXT_DUP); > if (rc == MDB_SUCCESS) { > if (mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { > MDB_cursor *mx; > fetchm: > mx = &mc->mc_xcursor->mx_cursor; > data->mv_size = NUMKEYS(mx->mc_pg[mx->mc_top]) * > mx->mc_db->md_pad; > data->mv_data = METADATA(mx->mc_pg[mx->mc_top]); > mx->mc_ki[mx->mc_top] = NUMKEYS(mx->mc_pg[mx->mc_top])-1; ^^^ this line advances the cursor. > } else { > rc = MDB_NOTFOUND; > } > } > break; > > > On Tue, Aug 28, 2018 at 8:26 PM Howard Chu <openldap-its@openldap.org> > wrote: > >>> Full_Name: Karl Waclawek >>> Version: LMDB 0.9.22 >>> OS: Windows 10 >>> URL: ftp://ftp.openldap.org/incoming/ >>> Submission from: (NULL) (2607:fea8:7a80:814:3ca9:e8f6:801c:d6f8) >>> >>> >>> The documentation in lmdb.h for MDB_GET_MULTIPLE, MDB_NEXT_MULTIPLE and >>> MDB_PREV_MULTIPLE seems to have two issues: >>> >>> 1) The documentation indicates that both, the key and data, are >> returned, but >> in >>> reality only the data is returned. >> >> Looks like you're right. Since these are DUPs the key would be the same >> every time so there's no need to return it on each call. >> Doc fixed in git mdb.master. >> >>> 2) The documentation states that MDB_GET_MULTIPLE moves the cursor. This >> does >>> not seem to be true. >> >> The doc is correct here. The cursor is advanced far enough that a >> MDB_NEXT_MULTIPLE will return the following data, as opposed to just >> returning the same data again. >> > -- -- Howard Chu CTO, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc/ Chief Architect, OpenLDAP http://www.openldap.org/project/
Thanks! On Wed, Aug 29, 2018, 12:05 Howard Chu, <hyc@symas.com> wrote: > karl@waclawek.net wrote: > > --00000000000018b43a057492d87f > > Content-Type: text/plain; charset="UTF-8" > > > > As far as point 2 is concerned I was basing it on this piece of code, > where > > it looks like MDB_GET_MULTIPLE just checks some flags and then goes to > the > > fetchm label where the code does nothing but retrieve data: > > You're not reading the code correctly. See below. > > > > case MDB_GET_MULTIPLE: > > if (data == NULL || !(mc->mc_flags & C_INITIALIZED)) { > > rc = EINVAL; > > break; > > } > > if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { > > rc = MDB_INCOMPATIBLE; > > break; > > } > > rc = MDB_SUCCESS; > > if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) || > > (mc->mc_xcursor->mx_cursor.mc_flags & C_EOF)) > > break; > > goto fetchm; > > case MDB_NEXT_MULTIPLE: > > if (data == NULL) { > > rc = EINVAL; > > break; > > } > > if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { > > rc = MDB_INCOMPATIBLE; > > break; > > } > > rc = mdb_cursor_next(mc, key, data, MDB_NEXT_DUP); > > if (rc == MDB_SUCCESS) { > > if (mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { > > MDB_cursor *mx; > > fetchm: > > mx = &mc->mc_xcursor->mx_cursor; > > data->mv_size = NUMKEYS(mx->mc_pg[mx->mc_top]) * > > mx->mc_db->md_pad; > > data->mv_data = METADATA(mx->mc_pg[mx->mc_top]); > > mx->mc_ki[mx->mc_top] = NUMKEYS(mx->mc_pg[mx->mc_top])-1; > > ^^^ this line advances the cursor. > > > } else { > > rc = MDB_NOTFOUND; > > } > > } > > break; > > > > > > On Tue, Aug 28, 2018 at 8:26 PM Howard Chu <openldap-its@openldap.org> > > wrote: > > > >>> Full_Name: Karl Waclawek > >>> Version: LMDB 0.9.22 > >>> OS: Windows 10 > >>> URL: ftp://ftp.openldap.org/incoming/ > >>> Submission from: (NULL) (2607:fea8:7a80:814:3ca9:e8f6:801c:d6f8) > >>> > >>> > >>> The documentation in lmdb.h for MDB_GET_MULTIPLE, MDB_NEXT_MULTIPLE and > >>> MDB_PREV_MULTIPLE seems to have two issues: > >>> > >>> 1) The documentation indicates that both, the key and data, are > >> returned, but > >> in > >>> reality only the data is returned. > >> > >> Looks like you're right. Since these are DUPs the key would be the same > >> every time so there's no need to return it on each call. > >> Doc fixed in git mdb.master. > >> > >>> 2) The documentation states that MDB_GET_MULTIPLE moves the cursor. > This > >> does > >>> not seem to be true. > >> > >> The doc is correct here. The cursor is advanced far enough that a > >> MDB_NEXT_MULTIPLE will return the following data, as opposed to just > >> returning the same data again. > >> > > > > > > -- > -- Howard Chu > CTO, Symas Corp. http://www.symas.com > Director, Highland Sun http://highlandsun.com/hyc/ > Chief Architect, OpenLDAP http://www.openldap.org/project/ >
fixed in mdb.master fixed in mdb.re09 (0.9.23)
changed notes changed state Test to Closed