From 1806c0c8f7aa4365f9f72c8ea51e947d1e93ccd9 Mon Sep 17 00:00:00 2001
From: Martin Steinegger <themartinsteinegger@gmail.com>
Date: Wed, 6 Dec 2023 15:44:53 +0900
Subject: [PATCH] Add padding module

---
 src/CommandDeclarations.h    |  1 +
 src/MMseqsBase.cpp           |  9 ++++++++-
 src/util/CMakeLists.txt      |  1 +
 src/util/makepaddedseqdb.cpp | 31 +++++++++++++++++++++++++++++++
 4 files changed, 41 insertions(+), 1 deletion(-)
 create mode 100644 src/util/makepaddedseqdb.cpp

diff --git a/src/CommandDeclarations.h b/src/CommandDeclarations.h
index 503663424..e8eac296c 100644
--- a/src/CommandDeclarations.h
+++ b/src/CommandDeclarations.h
@@ -23,6 +23,7 @@ extern int convertkb(int argc, const char **argv, const Command& command);
 extern int convertmsa(int argc, const char **argv, const Command& command);
 extern int convertprofiledb(int argc, const char **argv, const Command& command);
 extern int createdb(int argc, const char **argv, const Command& command);
+extern int makepaddedseqdb(int argc, const char **argv, const Command& command);
 extern int createindex(int argc, const char **argv, const Command& command);
 extern int createlinindex(int argc, const char **argv, const Command& command);
 extern int createseqfiledb(int argc, const char **argv, const Command& command);
diff --git a/src/MMseqsBase.cpp b/src/MMseqsBase.cpp
index a8b8a8422..8ec4493eb 100644
--- a/src/MMseqsBase.cpp
+++ b/src/MMseqsBase.cpp
@@ -130,6 +130,13 @@ std::vector<Command> baseCommands = {
                 "<i:fastaFile1[.gz|.bz2]> ... <i:fastaFileN[.gz|.bz2]>|<i:stdin> <o:sequenceDB>",
                 CITATION_MMSEQS2, {{"fast[a|q]File[.gz|bz2]|stdin", DbType::ACCESS_MODE_INPUT, DbType::NEED_DATA | DbType::VARIADIC, &DbValidator::flatfileStdinAndGeneric },
                                                            {"sequenceDB", DbType::ACCESS_MODE_OUTPUT, DbType::NEED_DATA, &DbValidator::flatfile }}},
+        {"makepaddedseqdb",               makepaddedseqdb,              &par.onlyverbosity,              COMMAND_HIDDEN,
+                "Generate a padded sequence DB",
+                "Generate a padded sequence DB",
+                "Martin Steinegger <martin.steinegger@snu.ac.kr>",
+                "<i:sequenceDB> <o:sequenceDB>",
+                CITATION_MMSEQS2, {{"sequenceDB", DbType::ACCESS_MODE_INPUT, DbType::NEED_DATA|DbType::NEED_HEADER, &DbValidator::sequenceDb },
+                                          {"sequenceIndexDB", DbType::ACCESS_MODE_OUTPUT, DbType::NEED_DATA, &DbValidator::sequenceDb }}},
         {"appenddbtoindex",      appenddbtoindex,      &par.appenddbtoindex,      COMMAND_HIDDEN,
                 NULL,
                 NULL,
@@ -137,7 +144,7 @@ std::vector<Command> baseCommands = {
                 "<i:DB1> ... <i:DBN> <o:DB>",
                 CITATION_MMSEQS2, {{"DB", DbType::ACCESS_MODE_INPUT, DbType::NEED_DATA | DbType::VARIADIC, &DbValidator::allDb },
                                    {"DB", DbType::ACCESS_MODE_INPUT, DbType::NEED_DATA, &DbValidator::allDb }}},
-       {"indexdb",               indexdb,              &par.indexdb,              COMMAND_HIDDEN,
+        {"indexdb",               indexdb,              &par.indexdb,              COMMAND_HIDDEN,
                 NULL,
                 NULL,
                 "Martin Steinegger <martin.steinegger@snu.ac.kr>",
diff --git a/src/util/CMakeLists.txt b/src/util/CMakeLists.txt
index 3b2638fbf..7fa7b0603 100644
--- a/src/util/CMakeLists.txt
+++ b/src/util/CMakeLists.txt
@@ -31,6 +31,7 @@ set(util_source_files
         util/filterdb.cpp
         util/gff2db.cpp
         util/renamedbkeys.cpp
+        util/makepaddedseqdb.cpp
         util/masksequence.cpp
         util/maskbygff.cpp
         util/mergeclusters.cpp
diff --git a/src/util/makepaddedseqdb.cpp b/src/util/makepaddedseqdb.cpp
new file mode 100644
index 000000000..de2224159
--- /dev/null
+++ b/src/util/makepaddedseqdb.cpp
@@ -0,0 +1,31 @@
+#include "Parameters.h"
+#include "DBReader.h"
+#include "DBWriter.h"
+#include "Debug.h"
+#include "Util.h"
+
+int makepaddedseqdb(int argc, const char **argv, const Command &command) {
+    Parameters &par = Parameters::getInstance();
+    par.parseParameters(argc, argv, command, true, 0, 0);
+    DBReader<unsigned int> dbr(par.db1.c_str(), par.db1Index.c_str(), 1,
+                               DBReader<unsigned int>::USE_INDEX | DBReader<unsigned int>::USE_DATA);
+    dbr.open(DBReader<unsigned int>::SORT_BY_LENGTH);
+    DBWriter writer(par.db2.c_str(), par.db2Index.c_str(), 1, false, dbr.getDbtype());
+    writer.open();
+    std::string result;
+    const int ALIGN = 4;
+    for (long id = dbr.getSize() - 1; id >= 0; id--) {
+        unsigned int key = dbr.getDbKey(id);
+        char *data = dbr.getData(id, 0);
+        size_t seqLen = dbr.getSeqLen(id);
+        const size_t sequencepadding = (seqLen % ALIGN == 0) ? 0 : ALIGN - seqLen % ALIGN;
+        result.append(data, seqLen);
+        result.append(sequencepadding, ' ');
+        writer.writeData(data, seqLen + sequencepadding, key, 0, false);
+    }
+    writer.close(true);
+    DBReader<unsigned int>::softlinkDb(par.db1, par.db2, DBFiles::SEQUENCE_ANCILLARY);
+
+    dbr.close();
+    return EXIT_SUCCESS;
+}
\ No newline at end of file