.TH MS_INTRO 3 2013/07/17
.SH NAME
ms_intro - Introduction to libmseed

.SH INTRODUCTION

The Mini-SEED library provides a framework for manipulation of SEED
data records including the unpacking and packing of data records.
Functionality is also included for managing waveform data as
continuous traces.  All structures of SEED 2.4 data records are
supported with the following exceptions: Blockette 2000 opaque data
which has an unknown data structure by definition and Blockette 405
which depends on full SEED (SEED including full ASCII headers) for a
full data description.

The primary purpose of the library is to hide the details of Mini-SEED
in order to allow rapid development of Mini-SEED reading/writing
software.  The framework allows everything from manipulation of
Mini-SEED on a record-by-record basis to reading of Mini-SEED into
continuous trace segments to packing of large continuous traces using
a record template.

Certain common tasks have, through library design, been streamlined,
for example: reading Mini-SEED records from a file, adding data from
unpacked records to a group of traces or packing a group of continuous
traces into Mini-SEED records.

The following data encoding formats are supported for both unpacking
and packing: ASCII, INT16, INT32, FLOAT32, FLOAT64, STEIM1 and STEIM2.
The INT and FLOAT encodings each have two versions for quantities with
a different number of bits of representation.  The STEIM decompression
produces 32-bit integers; likewise the compression routines require
32-bit integers as input.  The following data encoding formats are
supported for unpacking only: GEOSCOPE (24-bit, 16/3 and 16/4 gain
ranged), CDSN, SRO and DWWSSN.

.SH DATA RECORDS

A Mini-SEED record is represented in the library using the data
structure given below.  This structure is used for both unpacking and
packing of Mini-SEED records.  When unpacking with msr_unpack(3) this
structure is populated.  When packing with msr_pack(3) this structure
is used as a template for the resulting data records and as a source
of samples to be packed.

Blockettes following the fixed section of the header are contained in
the blockette chain of BlktLink structures.  Shortcut pointers to
commonly used blockettes are maintained for types 100, 1000 and 1001.

Many common header fields which are not easily accessible/usable in
the raw header are available directly from the structure.  When this
structure is used as a packing template, these common header fields
are packed into the appropriate place in the fixed section or
blockette.  As examples, the ASCII stream identifiers (network,
station, location and channel) are available as NULL terminated
strings, the start time is available as a high precision epoch time
(see ms_time(3)) and the sample rate is available as a double
precision floating point value.

The MSRecord data structure:

.nf
typedef struct MSRecord_s {
  char           *record;            /* Mini-SEED record */
  int32_t         reclen;            /* Length of Mini-SEED record */
  
  /* Pointers to SEED data record structures */
  struct fsdh_s      *fsdh;          /* Fixed Section of Data Header */
  struct BlktLink    *blkts;         /* Root of blockette chain */
  struct blkt_100_s  *Blkt100;       /* Blockette 100, if present */
  struct blkt_1000_s *Blkt1000;      /* Blockette 1000, if present */
  struct blkt_1001_s *Blkt1001;      /* Blockette 1001, if present */
  
  /* Common header fields in accessible form */
  int32_t         sequence_number;   /* SEED record sequence number */
  char            dataquality;       /* Data quality indicator */
  char            network[11];       /* Network designation */
  char            station[11];       /* Station designation */
  char            location[11];      /* Location designation */
  char            channel[11];       /* Channel designation */
  hptime_t        starttime;         /* Record start time */
  double          samprate;          /* Nominal sample rate (Hz) */
  int64_t         samplecnt;         /* Number of samples in record */
  int8_t          encoding;          /* Data encoding format */
  int8_t          byteorder;         /* Byte order of record */
  
  /* Data sample fields */
  void           *datasamples;       /* Data samples */
  int64_t         numsamples;        /* Number of data samples */
  char            sampletype;        /* Sample type code: a, i, f, d */

  /* Stream oriented state information */
  StreamState    *ststate;           /* Stream processing state information */
}
MSRecord;
.fi

.SS Explanation of fields

.IP record:
Pointer to the Mini-SEED record which was unpacked into the MSRecord.

.IP reclen:
When unpacking this is the record length in bytes of the record
pointed to by the 'record' pointer.  When packing this is the length
of records to pack.

.IP fsdh:
A pointer to the Fixed Section of the Data Header, all appropriate
multi-byte quantities are in host byte order.

.IP blkts:
The root of the blockette chain.  The chain is constructed from
linked BlktLink structures.  All appropriate multi-byte quantities
in the blockettes are in host byte order.  The msr_addblockette(3)
routine can be used to add blockettes to this chain.  The BlktLink
structure and SEED blockette structures are defined in libmseed.h.

.IP Blkt100:
.IP Blkt1000:
.IP Blkt1001:
Shortcut pointers to common blockettes in the blockette chain.  If a
given blockette does not exist in the blockette chain the shortcut
pointer will be 0.  If more than one of these blockette types exist
in the chain this pointer will point to the last one.

.IP sequence_number:
SEED record sequence number, should be between 0 and 999999.

.IP dataquality:
Data record and quality indicator, should be 'D', 'R', 'Q' or 'M'.

.IP network:
.IP station:
.IP location:
.IP channel:
SEED stream identifiers as a NULL terminated strings.

.IP starttime:
Record start time, the time of the first sample, as a high precision
epoch time (see ms_time(3)).  This time can be converted using the
various ms_hptime2<X> functions.

.IP samprate:
The sample rate in samples per second in double precision.  During
unpacking this value will be set to the sample rate given in the 100
blockette if it is present, otherwise the sample rate derived from
the factor and multiplier in the fixed section of the header.  In a
packing template this value will be used to derive a factor and
multiplier for the fixed section of the header and will be written
into 100 blockettes if any are in the blockette chain.

.IP samplecnt:
The sample count, i.e. number of data samples in the record.

.IP encoding:
The SEED data sample encoding format.  During packing this dictates
what format will be used to pack the data samples.  Supported packing
formats are 0 (DE_ASCII), 1 (DE_INT16), 3 (DE_INT32), 4 (DE_FLOAT32),
5 (DE_FLOAT64), 10 (DE_STEIM1) and 11 (DE_STEIM2).

.IP byteorder:
Byte order of multi-byte quantities in the record.  A value of 0
indicates little endian and a value of 1 indicates big endian.
During packing this dictates the byte order of the final records.

.IP datasamples:
A pointer to the unpacked data samples.  If no data samples were
unpacked this will be 0.  The 'numsamples' field indicates how many
samples are in this array and the 'sampletype' field indicates what
type of samples they are.

.IP numsamples:
The number of samples pointed to by the 'datasamples' pointer.

.IP sampletype:
The type of samples pointed to by the 'datasamples' pointer.
Supported types are 'a' (ASCII), 'i' (integer), 'f' (float) and 'd'
(double).  The size of each sample type in bytes is returned
by the get_samplesize(3) lookup routine.

.IP ststate:
Pointer to a StreamState struct used internally to track stream
oriented state variables.  Memory for this only allocated when needed.

.SH TRACES

The library includes two facilities to manage collections of
continuous trace segments, each represented by their top most
structure: MSTraceGroup and MSTraceList.  The MSTraceList facility is
a next generation version of the MSTraceGroup facility.  Whereas the
MSTraceGroup facility uses a single linked list of time segments the
MSTraceList facility is slightly more complex with two levels of
linked lists and common access pointers.  The advantages are that the
MSTraceList structure is faster to populate, especially when there are
many segments (gappy data), and the list is always maintained in a
sorted order.

.SH TRACE LISTS

MSTraceList data structures allow the grouping of MSTraceID structures
which are themselves the root of MSTraceSeg structures, see libmseed.h
as a reference to these structures.

.SH TRACE GROUPS

MSTraceGroup data structures allow the grouping of MSTrace structures.
While a MSTrace structure is normally used to hold trace information
and associated data samples it can also be used without data samples
as a means to keep trace of data coverage without actual samples.

Numerous routines are provided for basic management of MSTrace
structures, including the creation of new MSTrace structures, adding
data from Mini-SEED data structures to MSTrace structures, printing
trace information, etc.

The MSTraceGroup data structure acts as a very simple place to begin a
chain of MSTrace structures and keep track of the number of traces.

The MSTrace and MSTraceGroup data structures:

.nf
typedef struct MSTrace_s {
  char            network[11];     /* Network designation */
  char            station[11];     /* Station designation */
  char            location[11];    /* Location designation */
  char            channel[11];     /* Channel designation */
  char            dataquality;     /* Data quality indicator */
  char            type;            /* MSTrace type code */
  hptime_t        starttime;       /* Time of first sample */
  hptime_t        endtime;         /* Time of last sample */
  double          samprate;        /* Nominal sample rate (Hz) */
  int64_t         samplecnt;       /* Num. in trace coverage */
  void           *datasamples;     /* Data samples */
  int64_t         numsamples;      /* Num. samples in datasamples */
  char            sampletype;      /* Sample type code: a, i, f, d */
  void           *prvtptr          /* Private pointer for general use */
  struct MSTrace_s *next;          /* Pointer to next trace */
}
MSTrace;

typedef struct MSTraceGroup_s {
  int32_t           numtraces;     /* Number of MSTraces in trace chain */
  struct MSTrace_s *traces;        /* Root of the trace chain */
}
MSTraceGroup;
.fi

.SS Explanation of fields

.IP dataquality:

SEED data quality indicator, either 'D', 'R', 'Q' or 'M'.  This value
will be (binary) 0 when the quality is unknown or mixed.

.IP network:
.IP station:
.IP location:
.IP channel:
MSTrace identifiers as a NULL terminated strings.

.IP type:
A single character trace type indicator.  This field is not used by
libmseed but could be used for application specific trace
identification.

.IP starttime:
MSTrace start time, the time of the first sample, as a high precision
epoch time (see ms_time(3)).  This time can be converted using the
various ms_hptime2<X> functions.

.IP endtime:
MSTrace end time, the time of the last sample, as a high precision epoch
time (see ms_time(3)).  This time can be converted using the
various ms_hptime2<X> functions.

.IP samprate:
The sample rate in samples per second in double precision.

.IP samplecnt:
The sample count, i.e. number of data samples in the trace.

.IP datasamples:
A pointer to the data samples.  If no data samples are included this
will be 0.  The 'numsamples' field indicates how many samples are in
this array and the 'sampletype' field indicates what type of samples
they are.

.IP numsamples:
The number of samples pointed to by the 'datasamples' pointer.

.IP sampletype:
The type of samples pointed to by the 'datasamples' pointer.
Supported types are 'a' (ASCII), 'i' (integer), 'f' (float) and 'd'
(double).  The size of each sample type in bytes is returned
by the get_samplesize(3) lookup routine.

.IP prvtptr:
A private pointer for general use.  This pointer is not used by
libmseed and can safely be used by the calling program.

.IP ststate:
Pointer to a StreamState struct used internally to track stream
oriented state variables.  Memory for this only allocated when needed.

.IP next:
A pointer to the next MSTrace structure.  The value will be 0 for the
last link in a chain of MSTrace structures.

.SH LOG MESSAGES

All of the log and diagnostic messages emitted by the library
functions use the same interface.  The output from this interface can
be controlled.  This is useful when the library will be embedded in a
larger system with a custom logging facility.  See the man page for
more details.

.nf
  ms_log() : the central logging facility.  Behavior is controlled by
        the settings specified with ms_loginit().

  ms_loginit() : set the functions and prefixes used for log,
        diagnostic and error messages.
.fi

The default destination for log messages is standard output (stdout),
while all diagnostic (including error) messages go to standard error
(stderr).  Most of the internal messages emmited by the library are
considered diagnostic and will, by default, go to standard error.

The default prefix for log and diagnostic messages is nothing. The
default prefix for diagnostic error messages is "Error: ".

There are reentrant versions of these functions that operate directly
on a logging parameter MSLogParam struct.  These are intended for use
in threaded programs or where a complex logging scheme is desired.
See the man pages for more details.


.SH WAVEFORM DATA

Waveform data samples are managed by libmseed in a couple of different
formats depending on how they are unpacked or will be packed.  An
array of samples is completely represented by an array of sample
values, the number of samples and a sample type.  The number of
samples is always the actual number of sample values, not the number
of bytes needed for storing the values.  Samples can be either ASCII,
32-bit integer, 32-bit floats or 64-bit double precision floats.

Sample types are identified by a single ASCII type character:
.nf
"a" - ASCII (8 bits)
"i" - integer (32 bits)
"f" - float (32 bits)
"d" - double (64 bits)
.fi

The size of each sample type in bytes is returned by the
get_samplesize(3) lookup routine.

.SH BYTE ORDER

The SEED 2.4 standard allows data only SEED (Mini-SEED) to be either
in big (most significant byte first) or little (least significant byte
first) endian byte order.  Unfortunately it is not well defined what
little endian Mini-SEED really means.  While libmseed supports all
four combinations of big and little endian header and data the surest
way to avoid compatibility problems is to always create big endian
Mini-SEED records (header and data).

.B Reading MiniSEED - how libmseed determines the byte order of a record:

The byte order of a record header is determined by checking if the
record start year is a sane value (e.g. between 1920 and 2020).  The
byte order of (compressed) data samples is determined by the byte
order flag in the Blockette 1000, if a Blockette 1000 is not present
the byte order is assumed to be the same as the header.  To force the
byte order determination of either the header or data section of a
record the following environment variables can be set:

.nf
UNPACK_HEADER_BYTEORDER
UNPACK_DATA_BYTEORDER
.fi

These variables should be set to either 0 (little endian) or 1 (big
endian).  A programmatic equivalent of setting these environment
variables is provided via the following macros:

.nf
MS_UNPACKHEADERBYTEORDER(X)
MS_UNPACKDATABYTEORDER(X)
.fi

.B Writing MiniSEED - in what byte order libmseed creates records:

Normally the byte order of MiniSEED created by libmseed is controlled
via a flag in the API.  This byte order flag determines the ordering
for both the header and data sections of a record.  To force the byte
order of either the header or data section of a record the following
environment variables can be set:

.nf
PACK_HEADER_BYTEORDER
PACK_DATA_BYTEORDER
.fi

These variables should be set to either 0 (little endian) or 1 (big
endian).  A programmatic equivalent of setting these environment
variables is provided via the following macros:

.nf
MS_PACKHEADERBYTEORDER(X)
MS_PACKDATABYTEORDER(X)
.fi

Note that some interpretations of the SEED 2.4 format imply that
so-called little endian MiniSEED means that the record header is
little endian but that the data section is big endian (as the only
defined data encodings must be based on the SEED DDL which, in turn,
must be defined in terms of big endian).  Libmseed will not create
MiniSEED of this flavor by default but can be configured to do so by
setting the environment variables described above appropriately.

.SH COMMON USAGE

Example programs using libmseed are provided in the 'examples'
directory of the source code distribution.

One of the most common tasks is to read a file of Mini-SEED records
and either perform some action based on the header values or apply
some process to the data samples.  This task is greatly simplified by
using the library functions ms_readmsr(3) and ms_readtraces(3).  The
ms_readmsr(3) routine will open a specified file and return MSRecord
structures for each Mini-SEED record it reads from the file.  The
ms_readtraces(3) routine will do the same except add all the data read
to a MSTraceGroup, this is ideal for quickly reading data for
processing.  Both of these routines are able to automatically detect
record length.

If your application is not designed to read Mini-SEED from files the
library also provides functions to detect and parse Mini-SEED records
in memory buffers.  For more information see ms_detect(3) and
msr_parse(3).

Skeleton code for reading a file with ms_readmsr(3):

.nf
main() {
  MSRecord *msr = NULL;
  int retcode;

  while ( (retcode = ms_readmsr (&msr, filename, 0, NULL, NULL, 1, 0, verbose)) == MS_NOERROR )
    {
       /* Do something with the record here, e.g. print */
       msr_print (msr, verbose);
    }

  if ( retcode != MS_ENDOFFILE )
    ms_log (2, "Cannot read %s: %s\n", filename, ms_errorstr(retcode));

  /* Cleanup memory and close file */
  ms_readmsr (&msr, NULL, 0, NULL, NULL, 0, 0);
}

.fi
For reading two files with ms_readtraces(3):
.nf

main() {
  MSTraceGroup *mstg = NULL;
  int retcode;

  retcode = ms_readtraces (&mstg, filename, 0, -1.0, -1.0, 0, 1, 0, verbose);

  if ( retcode != MS_ENDOFFILE )
    ms_log (2, "Cannot read %s: %s\n", filename, ms_errorstr(retcode));

  retcode = ms_readtraces (&mstg, filename2, 0, -1.0, -1.0, 0, 1, 0, verbose);

  if ( retcode != MS_ENDOFFILE )
    ms_log (2, "Cannot read %s: %s\n", filename2, ms_errorstr(retcode));

  if ( ! mstg )
    {
      fprintf (stderr, "Error reading file\\n");
      return -1;
    }

  /* Do something with the traces here, e.g. print */
  mst_printtracelist (mstg, 0, verbose, 0);

  mst_freegroup (&mstg);
}
.fi

Another common task is to create (pack) Mini-SEED records. The library
supports packing of Mini-SEED either from MSRecord structures, MSTrace
structures or MSTraceGroup collections using, respectively, msr_pack(3),
mst_pack(3) or mst_packgroup(3).  In each case the appropriate data
structure and parameters are provided to the routine along with a
function pointer to a routine that will be called each time a record
is complete and should be disposed of.

When packing Mini-SEED records the concept of a record header template
is used, the template is always in the form of a MSRecord structure.
This allows the calling program to dictate the contents, with a few
exceptions, of the header in the final data records.

Skeleton code for creating (packing) Mini-SEED records with
mst_pack(3):

.nf
static void record_handler (char *record, int reclen, void *srcname) {
  if ( fwrite(record, reclen, 1, outfile) != 1 )
    {
      ms_log (2, "Error writing %s to output file\n", (char *)srcname);
    }
}

main() {
  int64_t psamples;
  int precords;
  MSTrace *mst;
  char srcname[50];

  mst = mst_init (NULL);

  /* Populate MSTrace values */
  strcpy (mst->network, "XX");
  strcpy (mst->station, "TEST");
  strcpy (mst->channel, "BHE");
  mst->starttime = ms_seedtimestr2hptime ("2004,350,00:00:00.000000");
  mst->samprate = 40.0;

  /* The datasamples pointer and numsamples counter will be adjusted by
     the packing routine, the datasamples array must be dynamic memory
     allocated by the malloc() family of routines. */
  mst->datasamples = dataptr; /* pointer to 32-bit integer data samples */  
  mst->numsamples = 1234;
  mst->sampletype = 'i';      /* declare type to be 32-bit integers */

  mst_srcname (mst, srcname, 0);

  /* Pack 4096 byte, big-endian records, using Steim-2 compression */
  precords = mst_pack (mst, &record_handler, srcname, 4096, DE_STEIM2,
                       1, &psamples, 1, verbose, NULL);

  ms_log (0, "Packed %"PRId64" samples into %d records\n",
             psamples, precords);

  /* Disconnect datasamples pointer, otherwise mst_free() will free it */
  mst->datasamples = NULL;

  mst_free (&mst);
}
.fi

.SH SEE ALSO
\fBmsr_unpack(3)\fP, \fBms_time(3)\fP and \fBmsr_pack(3)\fP

.SH AUTHOR
.nf
Chad Trabant
IRIS Data Management Center
.fi