Added an updated version of the CSV generator.

tools/lc2csv.py

+#!/usr/bin/env python
+
+import sys
+import labcomm
+
+
+class Reader(object):
+    def __init__(self, file):
+        self.file = open(file)
+
+    def read(self, count):
+        data = self.file.read(count)
+        if len(data) == 0:
+            raise EOFError()
+        return data
+
+    def mark(self, value, decl):
+        pass
+
+
+def flatten(sample, _type):
+    if isinstance(_type, labcomm.sample):
+        flatten(sample, _type.decl)
+    elif isinstance(_type, labcomm.array):
+        for e in sample:
+            flatten(e, _type.decl)
+    elif isinstance(_type, labcomm.struct):
+        for name, decl in _type.field:
+            flatten(sample[name], decl)
+    elif isinstance(_type, labcomm.BOOLEAN):
+        print "%d," % sample,
+    elif isinstance(_type, labcomm.STRING):
+        print "\"%s\"," % sample,
+    elif isinstance(_type, labcomm.primitive):
+        print "%s," % sample,
+    else:
+        print sample, _type
+
+
+def flatten_labels(sample, _type, prefix=""):
+    if isinstance(_type, labcomm.sample):
+        flatten_labels(sample, _type.decl, _type.name)
+    elif isinstance(_type, labcomm.array):
+        if len(_type.indices) != 1:
+            raise Exception("Fix multidimensional arrays")
+        if len(sample) == 0:
+            raise Exception("We dont't handle dynamical sizes yet %s" % _type)
+        for i in range(0, len(sample)):
+            flatten_labels(sample[i], _type.decl, prefix + "[%d]" % i)
+    elif isinstance(_type, labcomm.struct):
+        for name, decl in _type.field:
+            flatten_labels(sample[name], decl,
+                           prefix + "." + name)
+    elif isinstance(_type, labcomm.primitive):
+        print '"%s",' % prefix,
+    else:
+        print sample, _type
+
+
+def dump(sample, _type):
+    for k in sorted(_type.keys()):
+        flatten(sample[k], _type[k])
+    print
+
+
+def dump_labels(current, _type):
+    for k in sorted(_type.keys()):
+        flatten_labels(current[k], _type[k])
+    print
+
+
+def main():
+    d = labcomm.Decoder(Reader(sys.argv[1]))
+    seen = {}
+    current = {}
+    _type = {}
+    while True:
+        try:
+            (o, t) = d.decode()
+        except EOFError:
+            break
+        if o == None:
+            # Type declaration
+	    seen[t.name] = 0
+	    _type[t.name] = t
+        else:
+            current[t.name] = o
+            if len(current) == len(_type):
+                # We have got one sample of each
+                break
+    dump_labels(current, _type)
+    n = 0
+    while o != None:
+        if seen[t.name] > n:
+            n = seen[t.name]
+            dump(current, _type)
+        current[t.name] = o
+        seen[t.name] += 1
+        try:
+            (o, t) = d.decode()
+        except EOFError:
+            break
+    dump(current, _type)
+
+
+if __name__ == "__main__":
+    main()

[Anders Blomdell]

An encoder is allowed to add new types after sending of samples has started, maybe the code should/could handle such a case as well?

[Tommy Olofsson]

I thought that was a bit weird as well. I'll fix that.

[Tommy Olofsson]

It would be impractical when using it "live" during an experiment. Would it not?

[Anders Blomdell]

Not necessarily, could be useful for dynamicvv logging. If used for control, one has to make sure that registration has finished (with retransmissions, etc) before switching in time-critical things using the new types. I.e. not CS101 stuff, but I can see applications for it.

[Tommy Olofsson]

The current version tries parsing the whole file looking for registrations before outputting the header and then data. This got nowhere when following a log file... If a valid compromize would be to accept all registrations that are written when the application starts, and then continue following the file for changes, then there will be a merge request in the works.

Feel free to deny it if there is a better way.