Using fscanf to correctly store them in a struct and go to the next patient

I am trying to read a file using fscanf with a while loop: 1) name 2) surname 3) age 4) unit 5) Random_Numer 6)Doctors my problem is how can i save them in structure using fscanf (example: patient(i).doctor) so it stops reading when it starts to read the new patient and stores it for example (day 1=doctor1 doctor2). ------------------------------------------------------------ 1)John 2)Noah 3)40 4)clinic1 5)8 6)(day1)Doctor1 \t 6)(day1)Doctor2 \t 6)(day1)Doctor3 \n 6)(day2)Doctor4 \t 6)(day2)Doctor5 \n

[dpb]

Show us the structure of the file itself -- would be simplest to just attach a representative section.

Almost certainly, using low-level i/o functions will be the hardest way to go at it, but need to see the actual file and the desired structure format to know the better way.

[Walter Roberson]

Doctor4    Doctor5
Oliver Alexander

What is the signal that Oliver Alexander is a new patient, and not the name of one or two doctors seen on day 3? Is the Num after the clinic name always exactly 3 more than the total number of doctor entries? Is the key that patient names never have tab in them but doctor id are tab separated, leading to the algorithm:

• if this line is empty, we must be at end of file
• if there is one word on this line it is a doctor id
• if there are two or more words on this line but no tabs, this is a patient name
• otherwise each tab separated element is a doctor id

This leaves open the question of whether a doctor id can have multiple words. If so then is there a rule such as doctor ids are always followed by tab even if they are the last or only one on the line?

[STYLIANOS CHRISTOFOROU]

Ex first line is doc1 doc2 doc3

second line is doc1 doc2

i want the to be save in a struc for the first day this patient visited doc1 and doc2 doc3 and second day visited doc1 doc2, when the name of the other patient starts how do i stop saving it into a stuct.

(i have a bit trouble explaining) ty for your answer

[STYLIANOS CHRISTOFOROU]

Yes Walter Roberson on the last questioned you asked

[STYLIANOS CHRISTOFOROU]

Each line of doctors represents the day

Line 1 day ) doc1 doc2 doc3

Line 2day ) doc1 doc2 doc3 doc4

[STYLIANOS CHRISTOFOROU]

It could be as many doctors entries Line 1 day) doc1 ... doc9

[dpb]

What it could be is of very little help -- if you want somebody to actually try to do something specific for your problem, post an example file...."help us help you!!!!"

[STYLIANOS CHRISTOFOROU]

File

[STYLIANOS CHRISTOFOROU]

John Noah <---Patien name p1)John Noah

40 <---Patient age p1)40

clinic1 <---Patient clinic p1)clinic1

8 <---Patient number p1)8

Doctor1 \t Doctor2 \t Doctor3\n <--p1)day1 with a visit doc1/doc2/doc3

Doctor4 \t Doctor5\n <--p1)day2 with a visit doc4/doc5

Oliver Alexander <---Patien name p2)Oliver Alexander

50 <---Patient age p2)50

clinic2 <---Patient clinic p2)clinic1

10 <---Patient number p2)10

Doctor1 \t Doctor2 \t Doctor3\n <--p2)day1 with a visit doc1/doc2/doc3

Doctor4 \t Doctor5 \t Doctor6\n <--p2)day2 with a visit doc4/doc5/doc6

Doctor7 <--p2)day3 with a visit doc7

[dpb]

Why don't you create the data file with some more structure to make it simpler to parse?

How will the input file be created in actual use instead of toy data?

Instead of forcing it all to be line-by-line parsing hoping to find a pattern that is recognizable; put some section headings in for a la Hansel & Gretel with the bread crumbs trail thru the forest...

While it would be possible (with a great deal of effort) to parse the above, why make it harder than needs be and likely fragile in real use when real names show up with all the variations they could have.

[STYLIANOS CHRISTOFOROU]

This is an assignment my teacher gaved me sorry for confusion.

[dpb]

What are the ground rules for the assignment? Surely not as nebulous as what you've described...

[STYLIANOS CHRISTOFOROU]

How can I explain to you the part you didn't get so it needs to scan the file for a name surname age clinic num for the first patient to be stored to the specific structure. For the days that he got visited by how many doctors and save them in a structure.

[dpb]

The part about whether this is the precise form REQUIRED by the assignment for the input file or whether designing an appropriate input file format to accomplish the objective is part of the assignment.

[STYLIANOS CHRISTOFOROU]

We can't edit the file it should be read as it is if understood correctly waht you asked

[dpb]

That's a terrible instructor, imo, then, if that is the precise input file given for the assignment.

[STYLIANOS CHRISTOFOROU]

It's a replica by me of what he gaved us by the way

[dpb]

That sucks...

I'd probably read the whole file into memory with readlines first, find the numeric fields to isolate sections and then process them.

What is the definition of the second number???? @Walter Roberson caught that at least for the two cases shown, it's the number of visits + 3 but your comments indicated it is a patient ID #? That the ID would be somehow correlated with the other number seems wrong if it really is an ID but more than strange if it isn't connected to that number.

Alternatively, the other route is that Walter was headed down -- read each line and parse it for content and make the decision of what to do with it; then go on to next line.

<GEEZER TALE ALERT>

This reminds me just a little of a sorry file format from the original plant computers of the B&W Oconee class reactors back 40+ years ago now. They were actually on punched paper tape which confounded things even more as the tapes would break, skip registration, etc., etc., etc., ... besides just the interpretation problems fo undefined fields only terminated by counting that was difficult to rely on for above reasons.

I ended up writing a routine there that read the whole tape, then built an input file from it that was parseable to actual data variables. It's probably overkill for a class assignment, but I'd be tempted to do something similar here when I found the sections -- although here with records, one can count the number between the successive patient name/age pair to get the number of visits and rely on the records being intact.

There, one only knew how many numbers were in a complete core power distribution output -- 177 fuel assemblies by 7 axial positions (one for each level of the incore detectors) and so could only go from how many one read and trying to find symmetry to locate where dropouts might have been from the actual number of values read -- and which were realistic and which were either nonsensical values or NaN if the reader skipped.

<END GEEZER>

[STYLIANOS CHRISTOFOROU]

 doctor id are tab separated random number of lines and random number of doctors in a line

[dpb]

But why is it always 3+ the number, then?

[STYLIANOS CHRISTOFOROU]

@dpb what do you mean 3+ the doctors? it is random lines of doctors and random tabs of doctors in a line.

[dpb]

It doesn't really matter -- as all the other discussion has shown, the file format can't be parsed by the initial definitions by trying to interpret each record sequentially based on its content -- not enough rules have been defined to make names unambiguous and the description of how the file is constructed is inconsistent with the actual content as I outlined in last exhaustive comment following Walter's regarding the naming conundrum.

It is possible that with only two samples, just random chance made the fact that this ID number turn out to be the same as the number of non-empty doctors in the two sections plus three; but it is a pretty small chance if the numbers were, indeed, both generated randomly. We don't know the range of values possible in the RNG, but even if it were just the range of 1:10 that covers the numbers here, that would only be a 1% chance (1 in 10 squared if independently sampled).

Can't you see the there are

Doctor1 -- Doctors5 --> 5 visits and ID --> 8 
Doctor1 -- Doctors7 --> 7 vists and ID --> 10

That simply doesn't appear to be random but that either the number of visits was set to be three fewer than the ID number generated or, conversely, the ID was set to be three more than the number of visits.

Anything is possible by chance; just not likely.

As the other notes illustrated, however, it's only possible to read the file if you do something similar to what I did(*) -- simply count the number of records between the two discernible bread crumbs of the numeric values and then infer what the other fields must be by their positions in the file relative to those; not by trying to parse the content of the text records -- "there be dragons!" that leads only to frustration because it isn't unambiguous.

Again, if this is really the file given in a course assignment, it was a very bad job of preparing a problem.

(*) Or, of course, it the above correlation between number of visits and the patient ID does hold, then that's the other piece of information that could be used to disambiguate the file content -- it's possible to isolate the ID field; having that number one can then read visits until find three fewer non-empty names at which time can presume the next record will be a new patient. Of course, the miscounted number of delimiters in some records means one still has to make assumptions of what the content is supposed to be...

[STYLIANOS CHRISTOFOROU]

First day of visit with doc1 \t doc2\n Second day of visit with doc1 \t doc2 \t doc3 \t doc4\n

[dpb]

Which doesn't address the Q? about the correlation between ID and number at all...and, it's all immaterial anyway, as Walter shows, there's no way to parse the file unequivically from the content and the provided rules by which to do so.

Answer 1

[dpb]

My approach would begin something like

P=readlines('Patients_Data.txt');                   % read the file as string array
ixNumbers=find(matches(P,digitsPattern));           % find the numeric records -- age, ID
nPatients=numel(ixNumbers)/2;                       % there are two numeric records/patient
for i=1:nPatients                                   % so, for each patient set
  j=2*i-1;                                          % skip every other age for indexing
  S(i).Patient=P(ixNumbers(j)-1);
  S(i).Age=str2double(P(ixNumbers(j)));
  S(i).ID=str2double(P(ixNumbers(j)+2));
  S(i).Clinic=P(ixNumbers(j)+1);
end

That gives the patient info; the doctors/days can be read from the section after the ID before the next patient; left as "exercise for student".

For the above, this yielded--

>> S(1)
ans = 
  struct with fields:

    Patient: "John Noah"
        Age: 40
         ID: 8
     Clinic: "clinic1"
>> S(2)
ans = 
  struct with fields:

    Patient: "Oliver Alexander"
        Age: 50
         ID: 10
     Clinic: "clinic2"
>>

In my opinion, that's a lot simpler than trying to parse the line record-by-record.

It's going to be VERY fragile in practice; but so will the attempt the other way; it's just a form rife for error in entry or parsing.

[STYLIANOS CHRISTOFOROU]

Thank you!

Answer 2

[Walter Roberson]

That file format cannot be parsed. Unlike what you indicated when I asked, there is no tab on the line for the solo doctor Doctor7 and therefore you cannot reliably tell the difference between a single doctor visit and a patient name.

This ambiguity could be resolved if it could be guaranteed that patient names always had at least one embedded space, but that doctor names do not have any spaces.

Note that there are some countries where people only use a single name, and do not have a surname, not even for legal purposes (as far as I can tell).

[STYLIANOS CHRISTOFOROU]

Ok thank you aswell.

[dpb]

"That file format cannot be parsed."

>> P
P = 
  13×1 string array
    "John Noah"
    "40"
    "clinic1"
    "8"
    "Doctor1 →Doctor2 →Doctor3"
    "Doctor4→→Doctor5"
    "Oliver Alexander"
    "50"
    "clinic2"
    "10"
    "Doctor1→→Doctor2→→Doctor3"
    "Doctor4→→Doctor5→→Doctor6"
    "Doctor7"
>> 

Indeed, as pointed out to begin with, the assignment is extremely flawed...and the tab isn't a tab at all, but char(8594). The first group also has an empty assignment as written; whether there's some meaning in that or whether it's just another aberration in the sorry file is anybody's guess.

>> D1= ["Doctor1 →Doctor2 →Doctor3";
    "Doctor4→→Doctor5"];
>> any(D1{1}==9)        % no tabs to be seen anywhere...
ans =
  logical
   0
>> 
>> find(double(D1{1})>127)
ans =
     9    18
>> double(D1{1}(ans))
ans =
        8594        8594
>> >> char(ans(1))
ans =
    '→'
>> 
>> split(D1,char(8594))
ans = 
  2×3 string array
    "Doctor1 "    "Doctor2 "    "Doctor3"
    "Doctor4"     ""            "Doctor5"
>> 

The second group has double-delimiters for both first two lines, but none for third so it will fail w/o fixup...

>> D2=["Doctor1→→Doctor2→→Doctor3";
    "Doctor4→→Doctor5→→Doctor6";
    "Doctor7"]
D2 = 
  3×1 string array
    "Doctor1→→Doctor2→→Doctor3"
    "Doctor4→→Doctor5→→Doctor6"
    "Doctor7"
>> split(D2,char(8594))
Error using split
Element 3 of the text contains 0 delimiters while the previous elements have 4. All elements must contain the same number of delimiters. 
>> 

One would have to use patterns or handle the missing/mismatched delimiters in some other fashion than just a simple regular split() operation as a well-designed input format would allow.

The only way to parse the file is to assume the data are of a given type positionally by their relative location with respect to a given field; independent of the actual content of the various name formats or delimiters.

Whether this is really the exact file the instructor provided or not we don't know, but as given here it's a testament to how not to design an input file structure.

[STYLIANOS CHRISTOFOROU]

Ignore the arrows think of it like (day1)Doctor1\t(day1)Doctor2\n (day2)Doctor4\t(day2)Doctor5\n

We don't know how many lines there will be neither the amount of doctors per tab.

[dpb]

There isn't a single \t character in the file first off...forget that idea.

As shown above, you also can't split() the fields without special processing to handle the screwed-up form in which there aren't the same number of delimiters per record.

As I showed in the Answer, the only way to parse this file is to make the assumptions that there are the two numeric fields for each patient and then infer the content of the other records based on their position relative to those.

As I showed (and left as "exercise for Student" since this is, after all, apparently a homework/class assignment), you can determine from those positions the number of records that are in the file between the present patient ID record and the next patient AGE record -- since it is presumed there's only one patient record, then one can also find out how many doctor/day records there are for the present patient and simply save what is there for each patient; dealing with the inconsistent number of delimiters however one chooses.

The outline is there; but unless you make gross assumption and just count records, it is simply not possible to decode this file from content alone.

End of story.