40. The WriteF function

40.1. Introduction

.intro: This document describes the WriteF() function, which allows formatted output in a manner similar to printf() from the Standard C library, but allows the Memory Pool Manager (MPM) to operate in a freestanding environment (see design.mps.exec-env).

.background: The documents design.mps.exec-env and design.mps.lib describe the design of the library interface and the reason that it exists.

40.2. Design

.no-printf: There is no dependency on printf(). The MPM only depends on mps_io_fputc() and mps_io_fputs(), via the library interface (design.mps.lib), part of the plinth. This makes it much easier to deploy the MPS in a freestanding environment. This is achieved by implementing our own output routines.

.writef: Our output requirements are few, so the code is short. The only output function which should be used in the rest of the MPM is WriteF().

Res WriteF(mps_lib_FILE *stream, Count depth, ...)

If depth is greater than zero, then the first output character, and each output character after a newline in a format string, is preceded by depth spaces.

WriteF() expects a format string followed by zero or more items to insert into the output, followed by another format string, more items, and so on, and finally a NULL format string. For example:

res = WriteF(stream, depth,
             "Hello: $A\n", (WriteFA)address,
             "Spong: $U ($S)\n", (WriteFU)number, (WriteFS)string,
             NULL);
if (res != ResOK)
  return res;

This makes Describe() methods much easier to write. For example, BufferDescribe() contains the following code:

res = WriteF(stream, depth,
             "Buffer $P ($U) {\n",
             (WriteFP)buffer, (WriteFU)buffer->serial,
             "  class $P (\"$S\")\n",
             (WriteFP)buffer->class, (WriteFS)buffer->class->name,
             "  Arena $P\n",       (WriteFP)buffer->arena,
             "  Pool $P\n",        (WriteFP)buffer->pool,
             "  ", buffer->isMutator ? "Mutator" : "Internal", " Buffer\n",
             "  mode $C$C$C$C (TRANSITION, LOGGED, FLIPPED, ATTACHED)\n",
             (WriteFC)((buffer->mode & BufferModeTRANSITION) ? 't' : '_'),
             (WriteFC)((buffer->mode & BufferModeLOGGED)     ? 'l' : '_'),
             (WriteFC)((buffer->mode & BufferModeFLIPPED)    ? 'f' : '_'),
             (WriteFC)((buffer->mode & BufferModeATTACHED)   ? 'a' : '_'),
             "  fillSize $UKb\n",  (WriteFU)(buffer->fillSize / 1024),
             "  emptySize $UKb\n", (WriteFU)(buffer->emptySize / 1024),
             "  alignment $W\n",   (WriteFW)buffer->alignment,
             "  base $A\n",        (WriteFA)buffer->base,
             "  initAtFlip $A\n",  (WriteFA)buffer->initAtFlip,
             "  init $A\n",        (WriteFA)buffer->ap_s.init,
             "  alloc $A\n",       (WriteFA)buffer->ap_s.alloc,
             "  limit $A\n",       (WriteFA)buffer->ap_s.limit,
             "  poolLimit $A\n",   (WriteFA)buffer->poolLimit,
             "  alignment $W\n",   (WriteFW)buffer->alignment,
             "  rampCount $U\n",   (WriteFU)buffer->rampCount,
             NULL);
if (res != ResOK)
  return res;

.types: For each format $X that WriteF() supports, there is a type WriteFX defined in mpmtypes.h, which is the promoted version of that type. These types are provided both to ensure promotion and to avoid any confusion about what type should be used in a cast. It is easy to check the casts against the formats to ensure that they correspond.

.types.cast: Every argument to WriteF() must be cast, because in variable-length argument lists the “default argument promotion” rules apply and this could cause an argument to be read incorrectly on some platforms: for example on a 64-bit platform the $W format, which expects a 64-bit argument, is incompatible with a 32-bit unsigned argument, which will not be promoted to 64 bits by the default argument promotion rules. (Note that most of these casts are unnecessary, but requiring them all makes it easy to check that the necessary ones are all there.)

.types.future: It is possibly that this type set or similar may be used in future in some generalisation of varargs in the MPS.

.formats: The formats supported are as follows.

Code

Name

Type

Example rendering

$A

address

Addr

000000019EF60010

$P

pointer

void *

000000019EF60100

$F

function

void (*)(void)

0001D69E01000000 (see .function)

$S

string

char *

hello

$C

character

char

x

$W

word

ULongest

0000000000109AE0

$U

decimal

ULongest

42

$B

binary

ULongest

00000000000000001011011110010001

$$

dollar

$

Note that WriteFC is an int, because that is the default promotion of a char (see .types).

.snazzy: We should resist the temptation to make WriteF() an incredible snazzy output engine. We only need it for Describe() methods. At the moment it’s a simple bit of code – let’s keep it that way.

.function: The F code is used for function pointers. The C standard defines conversion between pointer-to-function types (§6.3.2.3.8), but it does not define conversion between pointers to functions and pointers to data. To work around this, the bytes of their representation are written sequentially, and may have a different endianness to other pointer types. This output could be smarter, or even look up function names, but see .snazzy. The particular type void (*)(void) is chosen because in GCC (version 8) this suppresses the warning that we would otherwise get from -Wcast-function-type. See job004156_ and GCC Warning Options.