Smalltalk/X Webserver

Documentation of class 'ExternalStream':

Class: ExternalStream

Inheritance
Description
Related information
Class protocol
Instance protocol
Examples

Inheritance:

   Object
   |
   +--Stream
      |
      +--PeekableStream
         |
         +--PositionableStream
            |
            +--WriteStream
               |
               +--ReadWriteStream
                  |
                  +--ExternalStream
                     |
                     +--ExternalReadStream
                     |
                     +--ExternalWriteStream
                     |
                     +--FileStream
                     |
                     +--NonPositionableExternalStream

Package:: stx:libbasic

Category:: Streams-External

Version:: rev: 1.530 date: 2019/08/02 08:18:36; user: cg; file: ExternalStream.st directory: libbasic; module: stx stc-classLibrary: libbasic

Author:: Claus Gittinger; Stefan Vogel (many, many fixes ...)

Description:

ExternalStream defines protocol common to Streams which have a file-descriptor and
represent some file or communicationChannel of the underlying OperatingSystem.
ExternalStream is abstract; concrete classes are FileStream, PipeStream etc.

ExternalStreams can be in two modes: text- (the default) and binary-mode.
In text-mode, the elements read/written are characters;
while in binary-mode the basic elements are bytes which read/write as SmallIntegers
in the range 0..255.

Also, the stream can be either in buffered or unbuffered mode. In buffered mode,
data is not written until either a cr is written (in text mode) or a synchronizeOutput
is sent (in both modes).

The underlying OperatingSystem streams may either be closed explicitely (sending a close)
or just forgotten - in this case, the garbage collector will eventually collect the
object AND a close will be performed automatically (but you will NOT know when this
happens - so it is recommended, that you close your files when no longer needed).
Closing is also suggested, since if smalltalk is finished (be it by purpose, or due to
some crash) the data will not be in the file, if unclosed.
All streams understand the close message, so it never hurts to use it (it is defined as
a noop in one of the superclasses).

Most of the methods found here redefine inherited methods for better performance,
since I/O from/to files should be fast.

Recovering a snapshot:
  not all streams can be restored to the state they had before - see the implementation of
  reOpen in subclasses for more information.
  For streams sitting on some communication channel (i.e. Pipes and Sockets) you should
  reestablish the stream upon image restart (make someone dependent on ObjectMemory).
  FileStreams are reopened and positioned to their offset they had at snapshot time.
  This may fail, if the file was removed or renamed - or lead to confusion
  if the contents changed in the meantime.
  Therefore, it is a good idea to reopen files and check for these things at restart time.

[Instance variables:]

    handleType      <Symbol>        desribes what handle is:
                                        win32: #fileHandle, #socketHandle,
                                               #filePointer, #socketFilePointer, #pipeFilePointer
                                        unix: #filePointer, #socketFilePointer, #pipeFilePointer
                                    needed for win32, which uses different APIs for the different handles (sigh)
    handle          <Integer>       used to be always a filePointer somehow mapped to an integer (FILE* - not the fd);
                                    now, either a filePointer or a handle (win32)
    mode            <Symbol>        #readwrite, #readonly or #writeonly
    buffered        <Boolean>       true, if buffered (i.e. collects characters - does
                                    not output immediately)
    binary          <Boolean>       true if in binary mode (reads bytes instead of chars)
    eolMode         <Symbol>        one of nil, #nl, #cr or #crlf.
                                    determines how lines should be terminated.
                                    nil -> newLine (as in Unix);
                                    #crlf -> with cr-lf (as in MSDOS)
                                    #cr -> with cr (as in VMS)
    hitEOF          <Boolean>       true, if EOF was reached

    lastErrorNumber <Integer>       the value of errno (only valid right after the error -
                                    updated with next i/o operation)

[Class variables:]
    Lobby           <Registry>      keeps track of used ext-streams (to free up FILE*'s)

    StreamErrorSignal       <Signal> parent of all stream errors (see Stream class)
    InvalidReadSignal       <Signal> raised on read from writeonly stream
    InvalidWriteSignal      <Signal> raised on write to readonly stream
    InvalidModeSignal       <Signal> raised on text I/O with binary-stream
                                     or binary I/O with text-stream
    OpenErrorSignal         <Signal> raised if open fails
    StreamNotOpenSignal     <Signal> raised on I/O with non-open stream

Additional notes:
  This class is implemented using the underlying stdio-c library package, which
  has both advantages and disadvantages: since it is portable (posix defined), porting
  ST/X to non-Unix machines is simplified. The disadvantage is that the stdio library
  has big problems handling unbounded Streams, since the EOF handling in stdio is
  not prepared for data to arrive after EOF has been reached - time will show, if we need
  a complete rewrite for UnboundedStream ...

  Also, depending on the system, the stdio library behaves infriendly when signals
  occur while reading (for example, timer interrupts) - on real unixes (i.e. BSD) the signal
  is handled transparently - on SYS5.3 (i.e. non unixes :-) the read operation returns
  an error and errno is set to EINTR.
  That's what the ugly code around all getc-calls is for ...
  Since things get more and more ugly - we will rewrite ExternalStream
  completely, to NOT use any stdio stuff (and do its buffering itself).

  Notice that typical stdio's use a single errno global variable to return an error code,
  this was bad design in the stdio lib (right from the very beginning), since it's much
  harder to deal with this in the presence of lightweight processes, where errno gets
  overwritten by an I/O operation done in another thread. (stdio should have been written
  to return errno as a negative number ...).
  To deal with this, the scheduler treats errno like a per-thread private variable,
  and saves/restores the errno setting when switching to another thread.
  (Notice that some thread packages do this also, but ST/X's thread implementation
  does not depend on those, but instead uses a portable private package).

  Finally, if an stdio-stream is open for both reading and writing, we have to call
  fseek whenever we are about to read after write and vice versa.
  Two macros (__READING__ and __WRITING__) have been defined to be used before every
  fread/fgetc and fwrite/putc respectively.

Related information:

    FileStream
    Socket
    PipeStream
    Filename
    OperatingSystem

Class protocol:

Signal constants

inaccessibleSignal: ST-80 compatibility: return openErrorSignal
invalidModeSignal: return the signal raised when doing text-I/O with a binary stream
or binary-I/O with a text stream
invalidOperationSignal: return the signal raised when an unsupported or invalid
I/O operation is attempted
invalidReadSignal: return the signal raised when reading from writeonly streams
invalidWriteSignal: return the signal raised when writing to readonly streams
openErrorSignal: return the signal raised when a file open failed
streamIOErrorSignal: return the signal raised when an I/O error occurs.
(for example, a device-IO-error, or reading an NFS-dir,
which is no longer available and has been mounted soft)
streamNotOpenSignal: return the signal raised on I/O with closed streams

accessing

openStreams

answer a collection of open Streams having this class

usage example(s):

	self openStreams
	Socket openStreams
	NonPositionableExternalStream openStreams

defaults

defaultCopyBufferSize: anInteger: allows changing the size of buffer used when copying big
files/buld data from one stream to another.
Due to a bug on older Windows systems, the default used in copyToEndInto:
is chosen very conservatively, as the copy used to fail with big buffers.
However, modern Windows systems seem to not suffer from that bug,
and bigger buffer sizes will dramatically increase the performance of
copies to/from network drives.
This method allows for fine tuning, in case you run into performance problems,
or network buffer failures.
When nil, a somewhat conservative compromise is used.

error handling

errorReporter

I know about error codes

lastErrorNumber

return the errno of the last error

usage example(s):

     ExternalStream lastErrorNumber

lastErrorString

return a message string describing the last error

usage example(s):

     ExternalStream lastErrorString

reportOn: anErrorSymbolOrNumber

an error occurred.
Report it via an Exception

finalization

finalizationLobby: answer the registry used for finalization.
ExternalStreams have their own Registry

initialization

closeFiles: close all files.
To be called on exit of Smalltalk.
initDefaultEOLMode: unix EOL conventions
initModeStrings: initialize modeStrings which are passed down to the underlying
fopen/fdopen functions.
initialize: self patchByteOrderOptimizedMethods
patchByteOrderOptimizedMethods: EXPERIMENTAL (not yet done by default):
change the underlying implementation of
nextPutInt16MSB / nextPutInt16LSB
nextPutInt32MSB / nextPutInt32LSB
to the corresponding NATIVE methods.
reOpenFiles: reopen all files (if possible) after a snapShot load.
This is invoked via the #earlyRestart change notification.
update: something with: aParameter from: changedObject: have to reopen files when returning from snapshot

instance creation

forFileDescriptor: aFileDescriptor mode: modeSymbol

given a fileDescriptor, create an ExternalStream object
to operate on this fd.
The modeSymbol-argument is #readonly, #readwrite, ....
This may be used to wrap fd's as returned by user
primitive code, or to wrap pipe-fds into externalStreams.

usage example(s):

     the example below will probably fail (15 is a random FD):

     |s|

     s := ExternalStream forFileDescriptor:15 mode:'r'.
     s next.

forFileDescriptor: aFileDescriptor mode: modeSymbol buffered: buffered handleType: handleTypeSymbol

usage example(s):

     the example below will probably fail (15 is a random FD):

     |s|

     s := ExternalStream forFileDescriptor:15 mode:#readonly buffered:false handleType:#filePointer.
     s next.

forReadWriteToFileDescriptor: aFileDescriptor

given a fileDescriptor, create an ExternalStream object
to read/write from/to this fd. This may be used to wrap fd's
as returned by user primitive code, or to wrap pipe-
filedescriptors into externalStreams.

usage example(s):

     the example below will probably fail (15 is a random FD):

     |s|

     s := ExternalStream forReadWriteToFileDescriptor:15.
     s next.

forReadingFromFileDescriptor: aFileDescriptor

given a fileDescriptor, create an ExternalStream object
to read from this fd. This may be used to wrap fd's
as returned by user primitive code, or to wrap pipe-
filedescriptors into externalStreams.

usage example(s):

     the example below will probably fail (15 is a random FD):

     |s|

     s := ExternalStream forReadingFromFileDescriptor:15.
     s next.

forWritingToFileDescriptor: aFileDescriptor

given a fileDescriptor, create an ExternalStream object
to write to this fd. This may be used to wrap fd's
as returned by user primitive code, or to wrap pipe-
filedescriptors into externalStreams.

usage example(s):

     the example below will probably fail (15 is a random FD):

     |s|

     s := ExternalStream forWritingToFileDescriptor:15.
     s binary.
     s nextPut:1.

new

re-enable new - disabled in Stream superclass

obsolete

makePTYPair: obsolete since 12-07-2003

** This is an obsolete interface - do not use it (it may vanish in future versions) **
makePipe: obsolete since 12-07-2003

** This is an obsolete interface - do not use it (it may vanish in future versions) **

testing

isAbstract

Instance protocol:

Compatibility-Dolphin

beText

Compatibility-Squeak

nextInto: aByteArrayOrString: read elements into the argument, whose size determines the amount
of bytes to read. If not enough elements could be read, return
a copy of the argument; otherwise, return the filled argument.
readInto: aContainer startingAt: index count: nElements: same as #nextBytes:into:startingAt: for ByteArrays;
for LongArrays, nelements longs are read.
Squeak compatibility.
readOnly: Squeak compatibility: make the stream readOnly

Signal constants

invalidReadSignal
invalidWriteSignal

accessing

binary: switch to binary mode - default is text
binary: beBinaryBool: switch to binary mode - default is text
buffered: aBoolean: turn buffering on or off - default is on.
Answer the previous bufferd state.
contents: return the contents of the file from the current position up-to
the end. If the stream is in binary mode, a ByteArray containing
the byte values is returned.
In text-mode, a collection of strings, each representing one line,
is returned.
contentsAsString: to compensate for the bad naming, use this to make things explicit.
See also #contents, which returns the lines as stringCollection for textFiles.
contentsOfEntireFile: ST-80 compatibility: return contents as a String (or byteArray, if in binary mode).
See also #contents, which returns the lines as stringCollection for text files.
contentsSpecies: return the kind of object to be returned by sub-collection builders
(such as upTo)
dependents: (comment from inherited method)
return a Collection of dependents.
The default implementation here uses a global WeakDictionary to store
dependents
This may be too slow for high frequency change&update,
therefore, some classes (Model) redefine this for better performance.
Notice the mentioning of a WeakDictionary - read the classes documentation.
dependents: aDependentsCollection: (comment from inherited method)
set the collection of dependents.
The default implementation here uses a global Dictionary to store
dependents which may be too slow for high frequency change&update.
Therefore, some classes (Model) redefine this for better performance.
eolMode: return how end-of-line (EOL) is to be marked.
Returns one one of:
#crlf -> add a CR-NL, as in MSDOS
#cr -> add a CR, as in VMS
#nl -> add a NL, as in Unix
#eot -> add an EOT (= 0x04, as used in some modems/protocols)
#etx -> add an ETX (= 0x03, as used in some modems/protocols)
#nl -> add a NL, as in Unix
nil -> transparent
eolMode: aSymbolOrNil: specify how end-of-line (EOL) is to be marked.
The argument may be one of:
#crlf -> add a CR-NL, as in MSDOS
#cr -> add a CR, as in VMS
#nl -> add a NL, as in Unix
#eot -> add an EOT (= 0x04, as used in some modems/protocols)
#etx -> add an ETX (= 0x03, as used in some modems/protocols)
anyOther -> like #nl
fileDescriptor: return the fileDescriptor of the receiver -
notice: this one returns the underlying OSs fileDescriptor -
this may not be available on all platforms (i.e. non unix systems).
fileDescriptorOfFile: aHandle: for migration to rel5 only:
return the fileDescriptor of the argument handle -
notice: this one returns the underlying OSs fileDescriptor -
this may not be available on all platforms (i.e. non unix systems).
fileHandle: return the fileHandle of the receiver.
Under unix, this is the fileDescriptor; under windows, this is the Handle.
filePointer: return the filePointer of the receiver -
notice: for portability stdio is used; this means you will get
a FILE * - not a fileDescriptor.
(what you really get is a corresponding integer).
You cannot do much with the returned value
- except passing it to a primitive, for example.
handle
handleType
lineEndCRLF
lineEndLF
lineEndTransparent
pathName: answer the pathName of the stream.
Only FileStreams know the pathName, so we return an empty string here
readonly: set access mode to readonly
readwrite: set access mode to readwrite
text: switch to text mode - default is text
useCRLF: aBoolean: turn on or off CRLF sending (instead of LF only) - default is off.
This method is provided for backward compatibility - see #eolMode:
which offers another choice.

** This is an obsolete interface - do not use it (it may vanish in future versions) **
writeonly: set access mode to writeonly

closing

abortAndClose: close the stream - added for protocol compatibility with Socket and PipeStream.
see comment there
close: Close the stream.
No error if the stream is not open.
shutDown: close the stream - added for protocol compatibility with Socket.
see comment there

copying

copy: answer a copy of myself.
Have to dup the filedescriptor and position the copy

defaults

bufferSizeForBulkCopy: return the size of buffer used when copying big files/bulk data
from one stream to another.
Due to a bug on older Windows systems, the default used
is chosen very conservatively, as the copy used to fail with big buffers.
However, modern Windows systems seem to not suffer from that bug,
and bigger buffer sizes will dramatically increase the performance of
copies to/from network drives.
Setting DefaultCopyBufferSize allows for fine tuning,
in case you run into performance problems, or network buffer failures.
When nil, a somewhat conservative compromise is used.

error handling

argumentMustBeCharacter

report an error, that the argument must be a character in 0..FF

argumentMustBeInteger

report an error, that the argument must be an integer

argumentMustBeString

report an error, that the argument must be a string

errorAlreadyOpen

report an error, that the stream is already opened

usage example(s):

      self new errorAlreadyOpen

errorBinary

report an error, that the stream is in binary mode

errorNotBinary

report an error, that the stream is not in binary mode

errorNotBuffered

report an error, that the stream is not in buffered mode

errorReadOnly

report an error, that the stream is a readOnly stream

errorReporter

ST-80 mimicry.

errorUnsupportedOperation

report an error, that some unsupported operation was attempted

errorWriteOnly

report an error, that the stream is a writeOnly stream

ioError

report an error, that some I/O error occurred.
(for example, a device-IO-error, or reading an NFS-dir,
which is no longer available and has been mounted soft)

ioError: errorNumber

report an error, that some I/O error occurred.
(for example, a device-IO-error, or reading an NFS-dir,
which is no longer available and has been mounted soft)

lastErrorNumber

return the last error

lastErrorString

return a message string describing the last error

lastErrorSymbol

return an error symbol describing the last error

openError

report an error, that the open failed

openError: errorNumber

report an error, that the open failed

readError

report an error, that some read error occurred

readError: errorNumber

report an error, that some read error occurred

writeError

report an error, that some write error occurred

writeError: errorNumber

report an error, that some write error occurred

finalization

beExecutor: mark myself as being an executor for finalization.
We mis-use the mode symbol for this.
executor: return a copy for finalization-registration;
since all we need at finalization time is the fileDescriptor,
a cheaper copy is possible.
finalize: some Stream has been collected - close the file if not already done
reRegisterForFinalization
registerForFinalization: keep myself in newSpace, so it will be finalized early
unregisterForFinalization: probably the object will be collected soon...

initialization

initialize
initializeEOLMode

line reading/writing

nextLine: read the next line (characters up to newline).
Return a string containing those characters excluding the newline.
If the previous-to-last character is a cr, this is also removed,
so it's possible to read alien (i.e. ms-dos) text as well.
The line must be shorter than 32K characters - otherwise an error is signalled.
nextPutLine: aString: write the characters in aString and append an end-of-Line marker
(LF, CR or CRLF - depending in the setting of eolMode)
nextPutLinesFrom: aStream upToLineStartingWith: aStringOrNil: read from aStream up to and including a line starting with aStringOrNil
and append all lines to self.
Can be used to copy/create large files or copy from a pipe/socket.

If aStringOrNil is nil or not matched, copy proceeds to the end.
peekForLineStartingWith: aString: read ahead for next line starting with aString;
return the line-string if found, or nil if EOF is encountered.
If matched, do not advance position beyond that line
i.e. nextLine will read the matched line.
If not matched, reposition to original position for further reading.
peekForLineStartingWithAny: aCollectionOfStrings: read ahead for next line starting with any of aCollectionOfStrings;
return the index in aCollection if found, nil otherwise..
If no match, do not change position; otherwise advance right before the
matched line so that nextLine will return this line.

misc functions

async: aBoolean

set/clear the async attribute - if set, the availability of data on
the receiver will trigger an ioInterrupt.
If cleared (which is the default) no special notification is made.
Notice:
not every OS supports this
- check with OS>>supportsIOInterrupts before using

blocking: aBoolean

set/clear the blocking attribute - if set (which is the default)
a read (using next) on the receiver will block until data is available.
If cleared, a read operation will immediately return with a value of
nil.
Turning off blocking is useful when reading from PipeStreams
or Sockets, and the amount of data to be read is not known
in advance. However, the data must then be read using #nextBytes:
methods, in order to avoid other (pastEndRead) exceptions.

ioctl: ioctlNumber

to provide a simple ioctl facility - an ioctl is performed
on the underlying file; no arguments are passed.

ioctl: ioctlNumber with: arg

to provide a simple ioctl facility - an ioctl is performed
on the underlying file; the argument is passed as argument.
This is not used by ST/X, but provided for special situations
- for example, to control proprietrary I/O devices.

Since the type of the argument depends on the ioctl being
performed, different arg types are allowed here.
If the argument is nil, an ioctl without argument is performed.
If the argument is an integral number, it's directly passed;
if it's a kind of ByteArray (ByteArray, String or Structure),
or external data (ExternalBytes or ExternalAddress),
a pointer to the data is passed.
This allows performing most ioctls
- however, it might be tricky to setup the buffer.
Be careful in what you pass - ST/X cannot validate its correctness.

reset

set the read position to the beginning of the collection

setToEnd

redefined since it must be implemented differently

** This method raises an error - it must be redefined in concrete classes **

sync

make sure, that the OS writes cached data to the disk

syncData

make sure, that the OS writes cached data to the disk.
In this case, metadata is only written, if it is
required to read the file's data (so metadata will not be written,
if only access/modification time has changed).

tcgetattr

unix only:
provide the information of a tcgetattr call as a dictionary.
This provides terminal setting values of a tty or pty.
Used by terminal emulators.
On non-unix systems, an unsupported operation error is raised

usage example(s):

the system does not support tcgetattr

non homogenous reading

next: count into: aCollection startingAt: start

return the next count bytes or characters from the stream.

nextAvailable: count

return the next count elements of the stream as aCollection.
If the stream reaches the end before count elements have been read,
return what is available. (i.e. a shorter collection).
The type of collection is specified in #contentsSpecies.

usage example(s):

     (ReadStream on:#(1 2 3 4 5)) nextAvailable:3
     (ReadStream on:#(1 2 3 4 5)) nextAvailable:10
     (ReadStream on:'hello') nextAvailable:3
     (ReadStream on:'hello') nextAvailable:10

nextAvailable: count into: aCollection startingAt: pos

return the next count elements of the stream into aCollection at position pos.

nextAvailableBytes: count

return the next count bytes of the stream as a byteArray.
If the stream reaches the end before count bytes have been read,
return what is available. (i.e. a shorter byteArray).

nextAvailableBytes: count into: anObject startingAt: start

read the next count bytes into an object and return the number of
bytes read or the number of bytes read, if EOF is encountered before,
or no more bytes are available for reading (from the pipe/socket).

If the receiver is some socket/pipe-like stream, an exception
is raised if the connection is broken.

Notice, that in contrast to other methods,
this does NOT return nil on EOF, but the actual count.
Thus allowing read of partial blocks.

The object must have non-pointer indexed instvars
(i.e. it must be a ByteArray, String, Float- or DoubleArray).
If anObject is a string or byteArray and reused, this provides the
fastest possible physical I/O (since no new objects are allocated).

Use with care - non object oriented I/O.
Warning: in general, you cannot use this method to pass data from other
architectures since it does not care for byte order or float representation.

nextByte

read the next byte and return it as an Integer; return nil on error.
This is allowed in both text and binary modes, always returning the
bytes binary value as an integer in 0..255.

nextBytes: count into: anObject startingAt: start

read the next count bytes into an object and return the number of
bytes read or the number of bytes read, if EOF is encountered before.
If the receiver is some socket/pipe-like stream, an exception
is raised if the connection is broken.

Warning: if used with a pipe/socket, this blocks until the requested number
of bytes have been read. See #nextAvailableBytes:into:startingAt:
to only read whats there.

Notice, that in contrast to other methods,
this does NOT return nil on EOF, but the actual count.
Thus allowing read of partial blocks.

The object must have non-pointer indexed instvars
(i.e. it must be a ByteArray, String, Float- or DoubleArray),
or an externalBytes object (with known size).
If anObject is a string or byteArray and reused, this provides the
fastest possible physical I/O (since no new objects are allocated).

Use with care - non object oriented I/O.
Warning: in general, you cannot use this method to pass data from other
architectures (unless you prepared the buffer with care),
since it does not care for byte order or float representation.

nextInt16MSB: msbFlag

Read two bytes and return the value as a 16-bit signed Integer.
If msbFlag is true, value is read with most-significant byte first,
otherwise least-significant byte comes first.
A nil is returned if EOF is reached (also when EOF is hit after the first byte).
Works in both binary and text modes.

nextInt32MSB: msbFlag

Read four bytes and return the value as a 32-bit signed Integer,
which may be a LargeInteger.
If msbFlag is true, value is read with most-significant byte first,
otherwise least-significant byte comes first.
A nil is returned, if EOF is hit before all 4 bytes have been read.
Works in both binary and text modes.

nextUnsignedInt16MSB: msbFlag

Read two bytes and return the value as a 16-bit unsigned Integer.
If msbFlag is true, value is read with most-significant byte first,
otherwise least-significant byte comes first.
A nil is returned if EOF is reached (also when EOF is hit after the first byte).
Works in both binary and text modes.

nextUnsignedInt32MSB: msbFlag

Read four bytes and return the value as a 32-bit unsigned Integer, which may be
a LargeInteger.
If msbFlag is true, value is read with most-significant byte first, otherwise
least-significant byte comes first.
A nil is returned, if endOfFile occurs before all 4 bytes have been read.
Works in both binary and text modes.

nextWord

in text-mode:
read the alphaNumeric next word (i.e. up to non letter-or-digit).
return a string containing those characters.
in binary-mode:
read two bytes (msb-first) and return the value as a 16-bit
unsigned Integer (for compatibility with other smalltalks)

** This is an obsolete interface - do not use it (it may vanish in future versions) **

non homogenous writing

nextPutByte: aByteValue

write a byte.
Works in both binary and text modes.

nextPutInt16: anIntegerOrCharacter MSB: msbFlag

Write the argument, anIntegerOrCharacter as a short (two bytes).
If msbFlag is true, data is written most-significant byte first;
otherwise least first.
Notice that integers in the range 16r-8000 to +16rFFFF can be written
(i.e. both signed and unsigned int32 values can be written.
Works in both binary and text modes.

nextPutInt16LSB: anIntegerOrCharacter

do not remove.
See #patchByteOrderOptimizedMethods

nextPutInt16MSB: anIntegerOrCharacter

do not remove.
See #patchByteOrderOptimizedMethods

nextPutInt16NATIVE: anIntegerOrCharacter

Write the argument, anIntegerOrCharacter as a short (two bytes) in native byte order.
This is the CPU-specific byte order (LSB on x86 and VAX, MSB on sparc, and possibly on ARM).
Notice that integers in the range 16r-8000 to +16rFFFF can be written
(i.e. both signed and unsigned int32 values can be written.
Works in both binary and text modes.

nextPutInt32: aNumber MSB: msbFlag

Write the argument, aNumber as a long (four bytes).
If msbFlag is true, data is written most-significant byte first;
otherwise least first.
Notice that integers in the range 16r-80000000 to +16rFFFFFFFF can be written
(i.e. both signed and unsigned int32 values can be written.
Works in both binary and text modes.

nextPutInt32LSB: anIntegerOrCharacter

do not remove.
See #patchByteOrderOptimizedMethods

nextPutInt32MSB: anIntegerOrCharacter

do not remove.
See #patchByteOrderOptimizedMethods

nextPutInt32NATIVE: anInteger

Write the argument, anInteger as a long (four bytes) in native byte order.
This is the CPU-specific byte order (LSB on x86 and VAX, MSB on sparc and possibly on ARM).
Notice that integers in the range 16r-80000000 to +16rFFFFFFFF can be written
(i.e. both signed and unsigned int32 values can be written.
Works in both binary and text modes.
Notice: this message should not be sent explicitly by ANY program.
the following implementation replaces the code of either nextPutInt32MSB or LSB
dynamically (see #initialize on the class side)

nextPutUtf16: aCharacter

append my UTF-16 MSB representation to the argument, aStream.
UTF-16 can encode only characters with code points between 0 to 16r10FFFF.

usage example(s):

	(FileStream newTemporary
	    nextPutUtf16:$B;
	    nextPutUtf16:$Ä;
	    nextPutUtf16:(Character codePoint:16r10CCCC);
	    reset;
	    binary;
	    contents)

nextPutUtf16: aCharacter MSB: msbFlag

append my UTF-16 MSB representation to the argument, aStream.
UTF-16 can encode only characters with code points between 0 to 16r10FFFF.

positioning

position
position: index0Based

printing & storing

printOn: aStream

private

clearEOF
closeFile: low level close - may be redefined in subclasses
Don't send this message, send #close instead
closeFile: handle: for rel5 only
connectTo: aFileDescriptor withMode: modeSymbol handleType: handleTypeSymbol: connect a fileDescriptor; openmode is the string defining the way to open.
This can be used to connect an externally provided fileDescriptor (from
primitive code) or a pipeFileDescriptor (as returned by makePipe) to
a Stream object.
The openMode ('r', 'w' etc.) must match the mode in which
the fileDescriptor was originally opened (otherwise i/o errors will be reported later).
dupFd: duplicate my file descriptor
fopenMode: answer the mode for fopen.
Only used internally
open: aPath withMode: openModeString: low level open; opens the file/device and sets the handle instance
variable. Careful: this does not care for any other state.
reOpen: USERS WILL NEVER INVOKE THIS METHOD
sent after snapin to reopen streams.
cannot reopen here since I am abstract and have no device knowledge
setAccessor: what to: something: set the filePointer/fileHandle to the given one;
low level private & friend interface; may also be used to connect to some
externally provided handle.
setFileHandle: something: set the fileHandle to the given one;
low level private & friend interface; may also be used to connect to some
externally provided file handle.
setFileHandle: anIntegerOrExternalAddress mode: openMode: set the handle based upon a given fileHandle -
notice: this one is based on the underlying OSs fileDescriptor -
this is a fileDescriptor (Integer) on Unix and Windows, or an Handle (ExternalAddres) on Windows only.
It may not be available on all platforms.
setFilePointer: something: set the filePointer to the given one;
low level private & friend interface; may also be used to connect to some
externally provided file.
setLastError: aNumber
setSocketHandle: something: set the socketHandle to the given one;
low level private & friend interface; may also be used to connect to some
externally provided socket handle.

queries

nextError: return the error by trying to read something.
Should only be used, when we know, that a read operation
will return an error (otherwise a character may be lost).
Return an integer (error number), 0 (EOF) or nil (no error)
numAvailable

reading

next

return the next element; advance read position.
In binary mode, an integer is returned, otherwise a character.
If there are no more elements, either an exception is thrown or nil is returned
- see #pastEndRead.

next: count

return the next count elements of the stream as a collection.
Redefined to return a String or ByteArray instead of the default: Array.

nextOrNil

return the next element; advance read position.
In binary mode, an integer is returned, otherwise a character.
If there are no more elements, nil is returned.
This is #atEnd and #next in a single operation - to speed up some code

peek

return the element to be read next without advancing read position.
In binary mode, an integer is returned, otherwise a character.
If there are no more elements, either an exception is thrown or nil is returned
- see #pastEndRead.

peekOrNil

return the element to be read next without advancing read position.
In binary mode, an integer is returned, otherwise a character.
If there are no more elements, nil is returned.
Same as #atEnd and #peek in a single operation - speeding up some code

upToEnd

return a collection of the elements up-to the end.
Return an empty collection, if the stream-end is already at the end.

usage example(s):

     '/dev/null' asFilename readStream upToEnd
     '/proc/self/stat' asFilename readStream upToEnd
     'librun.so' asFilename readStream binary upToEnd

     self assert:('smalltalk.rc' asFilename readStream upToEnd size)
		  =  ('smalltalk.rc' asFilename readStream size)

rel5 protocol

atEndFile: handle: for migration to rel5 only
nextByteFromFile: handle: for migration to rel5 only
nextPutByte: aByte toFile: handle: for migration to rel5 only

testing

atEnd

return true, if position is at end.
If a stream is at the end, the next read operation
would return or raise an exception immediately without waiting.
If a stream is not at the end, this may be a blocking operation
(see NonPositionableExternalStream)

gotErrorOrEOF

answer true, if amn error or eof has been occurred on the stream

isBinary

return true, if the stream is in binary (as opposed to text-) mode.
The default when created is false.

isBlocking

return true, if O_NONBLOCK is NOT set in the fileDescriptor (probably UNIX specific)

isExecutor

am I an executor and no real stream object?

isExternalStream

return true, if the receiver is some kind of externalStream;
true is returned here - the method redefined from Object.

isOpen

return true, if this stream is open

usage example(s):

executors are only used to finalize - they are not really open and are closed with #closeFile,
     which does not check for #isOpen

isPipeStream

return true, if this stream is a pipe

isReadable

return true, if this stream can be read from

isSocket

return true, if this stream is a socket

isWritable

return true, if this stream can be written to

waiting for I/O

canBeSelected

return true, if this stream can be selected upon

numAvailableForRead

answer the number of bytes available for reading

usage example(s):

	'/etc/hosts' asFilename readStream numAvailableForRead

readWaitWithTimeoutMs: timeoutOrNil

suspend the current process, until the receiver
becomes ready for reading or a timeout (in milliseconds) expired.
If data is already available, return immediate.
Return true if a timeout occurred (i.e. false, if data is available).
The other threads are not affected by the wait.

readWriteWaitWithTimeoutMs: timeoutOrNil

suspend the current process, until the receiver
becomes ready for reading or writing or a timeout (in milliseconds) expired.
Return true if a timeout occurred (i.e. false, if data is available).
Return immediate if the receiver is already ready.
The other threads are not affected by the wait.

writeExceptionWaitWithTimeoutMs: timeoutOrNil

writeWaitWithTimeoutMs: timeoutOrNil

suspend the current process, until the receiver
becomes ready for writing or a timeout (in milliseconds) expired.
Return true if a timeout occurred (i.e. false, if data is available).
Return immediate if the receiver is already ready.
The other threads are not affected by the wait.

writing

cr: append an end-of-line character (or CRLF if in crlf mode).
reimplemented for speed
flush: write all buffered data - ignored if unbuffered
nextPut: aCharacter: write the argument, aCharacter - return nil if failed, self if ok.
Only single-byte characters are currently supported
nextPutAll: aCollection: write all elements of the argument, aCollection.
Reimplemented for speed when writing strings or byteArrays.
For others, falls back to general method in superclass.
nextPutAll: initialWriteCount from: buffer startingAt: initialOffset: append count elements with index starting at initialOffset
of the argument, buffer onto the receiver.
This is only allowed, if the receiver supports writing.
Answer the number of elements that were appended.
Buffer should be a byte collection.
I don't know how to write non-bytes to an external stream, but let superclass handle this.
nextPutAll: aCollection startingAt: start to: stop: write a range of elements of the argument, aCollection.
Reimplemented for speed when writing strings or byteArrays.
For others, falls back to general method in superclass.
nextPutAllUnicode: aString: normal streams can not handle multi-byte characters, so convert them to utf8.
This is needed, so that you can do ('something' asUnicode16String errorPrintCR)
nextPutBytes: count from: anObject startingAt: start: write count bytes from an object starting at index start.
return the number of bytes written - which could be 0.
The object must have non-pointer indexed instvars
(i.e. be a ByteArray, String, Float- or DoubleArray),
or an externalBytes object (with known size).

Use with care - non object oriented i/o.
Warning:
in general, you cannot use this method to pass non-byte data to other
architectures (unless you prepared the buffer with care),
since it does not care for byte order or float representation.
nextPutUnicode: aCharacter: normal streams can not handle multi-byte characters, so convert them to utf8

Examples:

open a file, read the contents and display it in a textView:

    |topView scrollPane textView fileStream text|

    topView := StandardSystemView new.
    topView label:'contents of Makefile'.

    scrollPane := HVScrollableView in:topView.
    scrollPane origin:0.0@0.0 corner:1.0@1.0.

    textView := EditTextView new.
    scrollPane scrolledView:textView.

    fileStream := 'Makefile' asFilename readStream.
    text := fileStream upToEnd.
    fileStream close.

    textView contents:text.

    topView open.

Notice, all of the above can also be done (simply) as:

    EditTextView openOn:'Makefile'

ST/X 7.2.0.0; WebServer 1.670 at bd0aa1f87cdd.unknown:8081; Mon, 04 Dec 2023 09:38:49 GMT