Tips and Tricks

I have many almost identical protocols

You can give arguments to a protocol. In the INP or OUT link, write:

field (OUT, "@protocolfile protocol(arg1,arg2,arg3) bus")

In the protocol, reference arguments as $1 $2 $3 or inside strings as "\$1 \$2 \$3".

moveaxis {out "move\$1 %.6f";}
field (OUT, "@motor.proto moveaxis(X) motor1")

readpressure {out 0x02 0x00 $1; in 0x82 0x00 $1 "%2r";}
field (INP, "@vacuumgauge.proto readpressure(0x84) gauge3")

I have a device that sends unsolicited data

Use I/O Intr processing. The record receives any input and processes only when the input matches.

read {in "new value = %f";}

record (ai, "$(RECORD)") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto read $(BUS)")
  field (SCAN, "I/O Intr")
}

I have a device that sends multi-line messages

Here is the value:
3.1415

Use as many in commands as you get input lines.

read_value {in "Here is the value:"; in "%f";}

I need to write more than one value in one message

There is more than one solution to this problem. Different approaches have different requirements.

A) All values have the same type and are separated by the same string

Use array records (e.g. waveform, aao).

array_out {separator=", "; out "an array: (%.2f)";}

The format %.2f is repeated for each element of the array. All elements are separated by ", ".
Output will look like this:

an array: (3.14, 17.30, -12.34)

B) We have up to 12 numeric values

Use a calcout record and redirection to fields.

write_ABC {out "A=%(A).2f B=%(B).6f C=%(C).0f";}

You must specify a valid expression in the CALC field even if you don't use it.

record (calcout, "$(RECORD)") {
  field (INPA, "$(A_RECORD)")
  field (INPB, "$(B_RECORD)")
  field (INPC, "$(C_RECORD)")
  field (CALC, "0")
  field (DTYP, "stream")
  field (OUT, "@$(DEVICETYPE).proto write_ABC $(BUS)")
}

C) Values are in other records on the same IOC

Use redirection to records.

acquire {out 'ACQUIRE "%(\$1:directory)s/%s",%(\$1:time).3f;';}

You can specify a record name or record.FIELD in parentheses directly after the %. To avoid plain record names in protocol files use protocol arguments like \$1. In the link, specify the record name or just the basename of the other records (device name) in parentheses.

record (stringout, "$(DEVICE):getimage") {
  field (DTYP, "stream")
  field (OUT, "@$(DEVICETYPE).proto acquire($(DEVICE)) $(BUS)")
}

I need to read more than one value from one message

Again, there is more than one solution to this problem.

A) All values have the same type and are separated by the same string

Use array records (e.g. waveform, aai).

array_in {separator=","; in "array = (%f)";}

The format %f is repeated for each element of the array. A "," is expected beween element.
Input may look like this:

array = (3.14, 17.30, -12.34)

B) The message and the values in it can be filtered easily

Use I/O Intr processing and value skipping (%*)

read_A {out "GET A,B"; in "A=%f, B=%*f";}
read_B {in "A=%*f, B=%f";}

record (ai, "$(DEVICE):A") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto read_A $(BUS)")
  field (SCAN, "1 second")
}
record (ai, "$(DEVICE):B") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto read_B $(BUS)")
  field (SCAN, "I/O Intr")
}

Record A actively requests values every second. The reply contains values A and B. Record A filters only value A from the input and ignores value B by using the * flag. Nevertheless, a complete syntax check is performed: B must be a valid floating point number. Record B is I/O Intr and gets (a copy of) any input, including input that was directed to record A. If it finds a matching string it ignores value A, reads value B and then processes. Any non-matching input is ignored by record B.

C) Values should be stored in other records on the same IOC

Use redirection to records. To avoid record names in protocol files, use protocol arguments.

read_AB {out "GET A,B"; in "A=%f, B=%(\$1)f";}

record (ai, "$(DEVICE):A") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto read_AB($(DEVICE):B) $(BUS)")
  field (SCAN, "1 second")
}
record (ai, "$(DEVICE):B") {
}

Whenever record A reads input, it stores the first value in its own VAL field as usual and the second in the VAL field of record B. Because the VAL field of record B has the PP attribute, this automatically processes record B.

I have a device that sends mixed data types: numbers or strings

Use a @mismatch exception handler and redirection to records. To avoid record names in protocol files, use protocol arguments.

Example

When asked "CURRENT?", the device send something like "CURRENT 3.24 A" or a message like "device switched off".

read_current {out "CURRENT?"; in "CURRENT %f A"; @mismatch {in "%(\$1)39c";}}

record (ai, "$(DEVICE):readcurrent") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto read_current($(DEVICE):message) $(BUS)")
}
record (stringin, "$(DEVICE):message") {
}

After processing the readcurrent record, you can see from SEVR/STAT if the read was successful or not. With some more records, you can clean the message record if SEVR is not INVALID.

record (calcout, "$(DEVICE):clean_1") {
  field (INPA, "$(DEVICE):readcurrent.SEVR CP")
  field (CALC, "A#3")
  field (OOPT, "When Non-zero")
  field (OUT, "$(DEVICE):clean_2.PROC")
}
record (stringout, "$(DEVICE):clean_2") {
  field (VAL, "OK")
  field (OUT, "$(DEVICE):message PP")
}

I need to read a web page

First you have to send a correctly formatted HTML request. Note that this request must contain the full URL like "http://server/page" and must be terminated with two newlines. The server should be the same as in the drvAsynIPPortConfigure command (if not using a http proxy). The web page you get often contains much more information than you need. Regular expressions are great to find what you are looking for.

Example 1

Read the title of a web page.

get_title {
  extrainput = ignore;
  replyTimeout = 1000;
  out "GET http://\$1\n\n";
  in "%+.1/(?im)<title>(.*)<\/title>/";
}

Terminate the request with two newlines, either explicit like here or using an outTerminator. The URI (without http:// but including the web server host name) is passed as argument 1 to \$1. Note that web servers may be slow, so allow some replyTimeout.

If you don't use an inTerminator then the whole page is read as one "line" to the in command and can be parsed easily with a regular expression. We want to see the string between <title> and </title>, so we put it into a subexpression in () and request the first subexpression with .1. Note that the / in the closing tag has be be escaped to avoid a misinterpretation as the closing / of the regular expression.

The tags may be upper or lower case like <TITLE> or <Title>, so we ask for case insensitive matching with (?i).

The string should be terminated with the first closing </title>, not the last one in the file. (There should not be more than one title but you never know.) Thus we ask not to be greedy with (?m). (?i) and (?m) can be combined to (?im). See the PCRE documentation for more regexp syntax.

The regular expression matcher ignores and discards any content before the matching section. Content after the match is discarded with extrainput = ignore so that it does not trigger errors reporting "surplus input".

Finally, the title may be too long for the record. The + tells the format matcher not to fail in this case but to truncate the string instead. You can read the string with a stringin record or for longer strings with a waveform record with data type CHAR.

record (stringin, "$(DEVICE):title") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto get_title($(PAGE)) $(BUS)")
}
record (waveform, "$(DEVICE):longtitle") {
  field (DTYP, "stream")
  field (INP, "@$(DEVICETYPE).proto get_title($(PAGE)) $(BUS)")
  field (FTVL, "CHAR")
  field (NELM, "100")
}

Example 2

Read a number from a web page. First we have to locate the number. For that we match against any known string right before the number (and discard the match with *). Then we read the number.

get_title {
  extrainput = ignore;
  replyTimeout = 1000;
  out "GET http://\$1\n\n";
  in "%*/Interesting value:/%f more text";
}

When using extrainput = ignore;, it is always a good idea to match a few bytes after the value, too. This catches errors where loading of the page is interrupted in the middle of the number. (You don't want to miss the exponent from something like 1.23E-14).

You can read more than one value from a file with successive regular expressions and redirections. But this only works if the order of the values is predictible. StreamDevice is not an XML parser! It always reads sequentially.