WRITE to a file in a specific location

[kpierce]

Hello:

I have some code that reads to a certain record in a text file, then extracts that record. That all works fine.

Now I need to OVERWRITE that same record once it is used.

How do I open the file as Output, read to a location then write over it? I hope this is possible without have to parse the file into temp files.

' This section works as designed. It goes to a random record in the file
Open GcmdTMKFile For Input As #1
Randomize
intRandom = Int((GintRecCount * Rnd) + 1)    ' Generate random value between 1 and TmpRecCount
' Loop through file until you get to the TmpRandom Record, then exit FOR NEXT LOOP
For X = 1 To intRandom
    Line Input #1, strRecord
Next
Close #1
strTMKRecord = strRecord

' Now I need to write "XXXXX" over the record I just extracted

Open GcmdTMKFile For Output As #1
For X = 1 To intRandom
    Line Input #1, strRecord   ' I get an error because I have opened the file for Output
Next
Print #1, "XXXXX"
Close #1

Any help is greatly appreciated.

thanks.
kpierce

[dglienna]

That would be BINARY MODE, which means one extra length title will overwrite the next track. Not a good idea.

Other approach, use TWO files, read one, and write the other.

[DataMiser]

It depends on what your file will be like. If the records are fixed lenght then you can use a random access file which would be the best and fastest way possible. With RA files you can read or write a specific record at will. With standard file methods you would need to open the file for input to read your data then open the file for output to write the data. To change a record in this type of file requires the file be rewritten so you either need to use a tmp file or read all the data into memory before attempting to write to the file.

Binary access is also an option in which you can seek to a specific location within the file and write data to it. This will cause issues if the data you are writing is not the same length as that which was there before. If you new data is shorter then some of the old record will still exist in the file unless of course you padd your new data with spaces to fill the space of the old record. Worse if the new data is longer then will will overwrite part of the data in the next record and corrupt your file. Proper planning must be done to use this type of file as well as the random access methods.

[kpierce]

Thanks to everyone for the help.

I will always know the length of a record, it will be fixed based on the file created and each record/row/field will be 4-9 characters in lengh.

I am not a wiz and am trying to simply apply a change without having to redesign file layouts or other areas of the code. I really appreciate all the help.

Let me give you an example:
TEST.TXT
1234
5678
2345
7654
6543

The first part of the code reads to a random 'record' number and loads that records value into a variable. For example the random record is 3, which contains 2345.

So the program now closes the file. (This is the old code, this is all working fine).

Now the change is I need to go back into record 3 and write over the value 2345 with ZZZZ.

So the file now becomes:
1234
5678
ZZZZ
7654
6543

I think I understand you are saying write the top of the file to a tempfile:
1234
5678

Then write ZZZZ

Then write the remainder of the file.

It seems like there should be a way I can get my location in the file when it is OPENed as OUTPUT, store that into a variable, example TmpLocation with the SEEK(1) function

Then CLOSE it and reOPEN it as BINARY/INPUT/RA, etc. and go directly to the position that was stored in TmpLocation.

The reason I want to try this way, is these files can become large (40Meg) in size and the program is already extremely slow.

Again, I really appreciate everyones support and guidance.

I have the following code to this point:

   ' Older code, reads a file and pulls out the required record. This file has CR+LF after each record, and the records are a fixed length in each file. all records within a given file are the same length.
    Open GcmdTMKFile For Input As #1
    Randomize
    intRandom = Int((GintRecCount * Rnd) + 1)    ' Generate random value between 1 and TmpRecCount
    ' Loop through file until you get to the TmpRandom Record, then exit FOR NEXT LOOP
    For X = 1 To intRandom
        Line Input #1, strRecord
    Next
    ' Find out the current location of the file.
    tmpLocator = Seek(1) ' I added this line to get the current location.
    Close #1
    strTMKRecord = strRecord

    ' **Need help with the following, please**
    ' I am adding this code. if the strTMKRecord starts with a "Z", it has already been used.
    ' Otherwise, I just want to write over the record I just selected with Line Input above.
    ' The following code inserts "ZZZZZZ", but it is in the middle of a record and it adds 10 bytes to the file.
    If Mid(strTMKRecord, 1, 1) <> "Z" Then
        ' If we are to this point we have a valid Record, and now need to overwrite the record in the file.
        Open GcmdTMKFile For Binary Access Read Write As #1
        Seek #1, tmpLocator - len(strTMKRecord) ' back up over the record.
        Put #1, , String(Len(strTMKRecord), "Z") & vbNewLine
        Close #1
    End If

Again, thank you all for all the help.
kp

[DataMiser]

When a file is opened for output a new file is created so there is no position to read. After you have written to a file that was opened for output you can check the lenght of the file which will tell you where you stopped writing but that is all.

When you open a file for Input this allows you to read from the file starting at the begining and each bit in order up until you are done. You can keep track of where you are by storing the number of bytes you have pulled during each read remembering of course to account for any CRLF combos if you are doing a line input.

The sample you showed looks like it is fixed lenght, 4 bytes of data + CRLF for a total of 6 bytes. If all records are to be this size then what you need to do is very simple but if the size varies you are limited as to what you can do. So anyway if every record (or row if you will) is the same length and that length is known then you can use Random Access rather than Input/Output methods.

For example if your file is 4 bytes per record with crlf for a total of 6 bytes you could search for and replace a value using random access with somethign like this.

Dim FileData As String * 6
Dim X As Long
Open TheFile For Random As #1 Len = 6
    Do While X < LOF(1) / 6
        Get #1, X, FileData
        If Mid(FileData, 1, 4) = SearchString Then
           FileData = "XXXX" & vbCrLf
            Put #1, X, FileData
            Exit Do
        End If
        X = X + 1
    Loop
Close #1

Note this is freehand code and has been a while since I had a need to do this type of file access so there may be a mistake or 2 in there but should give the basic idea.

ETA: As you can see this method would be much faster than searching a file then recreating the entire file especially if the search item is near the begining of the file.
The code above will read only until it finds the match then will replace the value and close the file without reading any more plus it will only be writing 6 bytes rather than the entire file. If the file were sorted then you could add additional code to speed up the search but if unsorted you pretty much have to read from the beging one row at a time until you find your match.

Of course if there could be more than one matching record you would have to remove the exit do statement so the code would continue to search to the end of the file for additional matches.

[kpierce]

Everyone has been a great help here, and thank you very much.

Based on the help, I now have this working. However, there is a new wrinkle that I need help working through.

There is a 'header' record in the file, I did not account for in my previous question.

The following code that was provided works great without the header. Can someone tell me how to account for this header? I tried starting with X=2, etc. but as the more advanced users here know, that simply advanced the record pointing to the next 9 characters in the file.

Ideally, I want to just disregard the top line. That is, start reading the file after the first CR+LF. The Header IS variable in length.

Any help is greatly appreciated. If I can figure that out, all the records are fixed length and this will work GREAT!

Thanks again in advance for all the help.
kp

Private Sub Command2_Click()
Dim FileData As String * 9
Dim X As Long
Dim tmpFile As String
tmpFile = App.Path & "\test32.txt"
Open tmpFile For Random As #1 Len = 9
strRecord = "8001426"
    X = 1
    Do While X < LOF(1) / 9
        Get #1, X, FileData
        If Mid(FileData, 1, 7) = strRecord Then
           FileData = "ZZZZZZZ" & vbCrLf
           Put #1, X, FileData
           Exit Do
        End If
        X = X + 1
    Loop
Close #1
End Sub

Example TEST32.TXT

6 0 6 9 1 4 2 1 0 0 
1001162
2001163
3001164
4001165
5001168
6001261
7001263
9001265
0001316
1001326
2001361
3001362
4001363
5001364
6001365
7001416
8001426
9001436
0001461
1001462
2001463
3001464
4001465

[dglienna]

Other approach, use TWO files, read one, and write the other.

You can easily split a file based on CR/LF, and eliminate the header.

Binary mode is not the best way to 'SKIN A CAT' (search for that)

[DataMiser]

To the best of my knowledge random access works only when all records ar eof the same size. If works extremely well when you know the record number you need to access. For searching a non sorted file without an index it is no better than using line input. For writing it is up there near the top as you only need to write a few bytes but must know the record number.

You could as suggested above read the file into memory and split it on the CRLF to get each line into an array element. You could then check the length of the first line, search the array then calculate the byte position within the original file and use binary access to write to the correct location. In some cases this would be slower [ such as if the item is within the first few records of a very large file ] in other cases it could be a lot faster [ record is near the end of a very large file ] and in some cases you may get a memory error [ file is to large to process this way ] The latter could be avoided by reading the file in chunks rather than trying to load the entire file at once.

The real question is why search a text file and replace a value like this, would be much better to use a database engine if possible.

If the entire file must be loaded and resaved you could just load it, call the replace() function then save.

The bottom line though is if you have records that are only 4-9 bytes in size that result in a 40+ meg text file and these records must be changed on a regular basis a text file is the worst way I could think of to handle it.