You've already forked mariadb-columnstore-engine
mirror of
https://github.com/mariadb-corporation/mariadb-columnstore-engine.git
synced 2025-07-24 22:42:05 +03:00
238 lines
5.2 KiB
C++
238 lines
5.2 KiB
C++
/* Copyright (C) 2014 InfiniDB, Inc.
|
|
|
|
This program is free software; you can redistribute it and/or
|
|
modify it under the terms of the GNU General Public License
|
|
as published by the Free Software Foundation; version 2 of
|
|
the License.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program; if not, write to the Free Software
|
|
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
|
|
MA 02110-1301, USA. */
|
|
|
|
/*
|
|
* A tool that can take a csv file of extent map entires and produce a BRM_saves_em-compatible file.
|
|
* If you re-compile extentmap.cpp to dump the extent map as it loads, you'll get a csv file on stdout.
|
|
* Save this to a file and edit it as needed (remove the cruft at the top & bottom for sure). Then use
|
|
* this tool to create a binary BRM_saves_em file.
|
|
*
|
|
* compile with
|
|
* g++ -g -Wall -o load_brm_from_file -I$HOME/genii/export/include -I/usr/include/libxml2
|
|
* load_brm_from_file.cpp
|
|
*
|
|
*/
|
|
#include <iostream>
|
|
#include <stdint.h>
|
|
#include <fstream>
|
|
#include <cerrno>
|
|
#include <string>
|
|
#include <cstdlib>
|
|
//#define NDEBUG
|
|
#include <cassert>
|
|
#include <limits>
|
|
using namespace std;
|
|
|
|
#include <boost/tokenizer.hpp>
|
|
using namespace boost;
|
|
|
|
#include "extentmap.h"
|
|
using namespace BRM;
|
|
|
|
#define EM_MAGIC_V4 0x76f78b1f
|
|
|
|
namespace BRM
|
|
{
|
|
EMEntry::EMEntry()
|
|
{
|
|
fileID = 0;
|
|
blockOffset = 0;
|
|
HWM = 0;
|
|
partitionNum = 0;
|
|
segmentNum = 0;
|
|
dbRoot = 0;
|
|
colWid = 0;
|
|
status = 0;
|
|
}
|
|
EMCasualPartition_struct::EMCasualPartition_struct()
|
|
{
|
|
lo_val = numeric_limits<int64_t>::min();
|
|
hi_val = numeric_limits<int64_t>::max();
|
|
sequenceNum = 0;
|
|
isValid = CP_INVALID;
|
|
}
|
|
} // namespace BRM
|
|
|
|
int main(int argc, char** argv)
|
|
{
|
|
int e;
|
|
|
|
int loadSize[3];
|
|
|
|
if (argc < 2)
|
|
{
|
|
cerr << "filename arg needed" << endl;
|
|
return 1;
|
|
}
|
|
|
|
ifstream in(argv[1]);
|
|
e = errno;
|
|
|
|
if (!in)
|
|
{
|
|
cerr << "file read error: " << strerror(e) << endl;
|
|
return 1;
|
|
}
|
|
|
|
// Brute force count the number of lines
|
|
int numEMEntries = 0;
|
|
|
|
string line;
|
|
|
|
getline(in, line);
|
|
|
|
while (!in.eof())
|
|
{
|
|
numEMEntries++;
|
|
getline(in, line);
|
|
}
|
|
|
|
// start at the beginning again...
|
|
in.clear();
|
|
in.seekg(0, ios_base::beg);
|
|
|
|
idbassert(in.good());
|
|
idbassert(in.tellg() == static_cast<streampos>(0));
|
|
|
|
string outname(argv[1]);
|
|
outname += ".out";
|
|
|
|
ofstream out(outname.c_str());
|
|
e = errno;
|
|
|
|
if (!out)
|
|
{
|
|
cerr << "file write error: " << strerror(e) << endl;
|
|
return 1;
|
|
}
|
|
|
|
loadSize[0] = EM_MAGIC_V4;
|
|
loadSize[1] = numEMEntries;
|
|
loadSize[2] = 1; // one free list entry
|
|
out.write((char*)&loadSize, (3 * sizeof(int)));
|
|
|
|
InlineLBIDRange fl;
|
|
fl.start = 0;
|
|
// the max lbid is 2^54-1, the size is in units of 1k
|
|
fl.size = numeric_limits<uint32_t>::max();
|
|
|
|
InlineLBIDRange maxLBIDinUse;
|
|
maxLBIDinUse.start = 0;
|
|
maxLBIDinUse.size = 0;
|
|
|
|
getline(in, line);
|
|
|
|
while (!in.eof())
|
|
{
|
|
EMEntry em;
|
|
int64_t v;
|
|
tokenizer<> tok(line);
|
|
tokenizer<>::iterator beg = tok.begin();
|
|
#if 0
|
|
emSrc[i].range.start
|
|
<< '\t' << emSrc[i].range.size
|
|
<< '\t' << emSrc[i].fileID
|
|
<< '\t' << emSrc[i].blockOffset
|
|
<< '\t' << emSrc[i].HWM
|
|
<< '\t' << emSrc[i].partitionNum
|
|
<< '\t' << emSrc[i].segmentNum
|
|
<< '\t' << emSrc[i].dbRoot
|
|
<< '\t' << emSrc[i].colWid
|
|
<< '\t' << emSrc[i].status
|
|
<< '\t' << emSrc[i].partition.cprange.hi_val
|
|
<< '\t' << emSrc[i].partition.cprange.lo_val
|
|
<< '\t' << emSrc[i].partition.cprange.sequenceNum
|
|
<< '\t' << (int)(emSrc[i].partition.cprange.isValid)
|
|
#endif
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.range.start = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.range.size = v;
|
|
|
|
if (em.range.start > maxLBIDinUse.start)
|
|
{
|
|
maxLBIDinUse.start = em.range.start;
|
|
maxLBIDinUse.size = em.range.size;
|
|
}
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.fileID = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.blockOffset = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.HWM = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.partitionNum = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.segmentNum = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.dbRoot = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.colWid = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.status = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.partition.cprange.hi_val = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.partition.cprange.lo_val = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.partition.cprange.sequenceNum = v;
|
|
|
|
v = strtoll(beg->c_str(), 0, 0);
|
|
++beg;
|
|
em.partition.cprange.isValid = v;
|
|
|
|
out.write((char*)&em, sizeof(em));
|
|
|
|
getline(in, line);
|
|
}
|
|
|
|
fl.start = maxLBIDinUse.start + maxLBIDinUse.size * 1024;
|
|
fl.size -= fl.start / 1024;
|
|
|
|
out.write((char*)&fl, sizeof(fl));
|
|
|
|
out.close();
|
|
in.close();
|
|
|
|
return 0;
|
|
}
|